Effect of clay colloids on radionuclide migration. Valtteri Suorsa

Transcription

1 Effect of clay colloids on radionuclide migration Valtteri Suorsa Master s thesis Supervisor Pirkko Hölttä Radiochemistry Department of Chemistry University of Helsinki Finland March 2017

2 Tiedekunta/Osasto Fakultet/Sektion Faculty Faculty of Science Laitos/Institution Department Department of Chemistry Tekijä/Författare Author Matti Valtteri Suorsa Työn nimi / Arbetets titel Title Effect of clay colloids on radionuclide migration Oppiaine /Läroämne Subject Radiochemistry Työn laji/arbetets art Level Master's Thesis Aika/Datum Month and year 03/2017 Sivumäärä/ Sidoantal Number of pages 118 Tiivistelmä/Referat Abstract In Finland, the spent nuclear fuel will be deposited at a depth of 400 m in the granitic bedrock. The disposal is based on KBS-3 concept, which relies on the multi-barrier principle, where different successive barriers prevent the migration of radionuclides to biosphere. The spent nuclear fuel is placed in the disposal tunnels in copper-iron canisters, which are surrounded by bentonite clay to insulate them from the groundwater flow and protect from the movements of the bedrock. Bentonite clay consists mainly of montmorillonite, which like the other aluminosilicates are known to retain radionuclides thus, contributing to the retention or immobilization of them. Besides the contribution to the multi-barrier system, the bentonite buffer is assumed to be a potential source of colloids due to the erosion of bentonite in certain conditions. Colloids in the context of radionuclide migration are nanoparticles in the size range from 1 to 1000 nm that remain suspended in water. The montmorillonite colloids could potentially act as carriers for otherwise immobile radionuclides like transuranium elements in the case of canister failure. Especially, 241 Am is an important radionuclide regarding the long-term safety of the final disposal as after a few hundred years 241 Am and its mother 241 Pu contribute most to the radiotoxicity of the spent nuclear fuel. The relevance of the colloids to the long-term performance is depending on several factors like colloid stability, mobility and their interaction with radionuclides. The colloid stability is depending on the groundwater conditions like ionic strength and ph. In low salinity groundwaters, the montmorillonite colloids have been shown to be stable. On the other hand, the collective processes of the rock matrix, bentonite colloids and radionuclides have to be investigated to assess the long-term performance of the multi-barrier system. It requires the combination of the different scale experiments from the simple laboratory experiments to large, natural scale in-situ experiments to understand the complex processes affecting the colloid-facilitated radionuclide migration. The large-scale laboratory experiments conducted with granite blocks offer an intermediate between the two extremes having a more natural system than the former and a better controllability than the latter. In this study, the radionuclide migration was studied in different scale laboratory experiments. The colloid-facilitated transport of Eu was studied with a block-scale experiment using a granite block with a natural water conducting fracture. The suitability of the block was assessed by conducting several experiments using different non-sorbing and sorbing tracer and montmorillonite colloids separated from synthetic Ni-labeled montmorillonite and Nanocor PGN Montmorillonite (98 %). Laser-induced breakdown detection (LIBD), photon correlation spectroscopy (PCS) and ICP-/MP-OES were utilized in colloid detection. Supportive batch experiments were conducted to study the colloid stability in different ground waters and the interaction between the granite, different montmorillonite colloids and Eu, an analog to Am. Good reproducibility was obtained with non-sorbing tracers. The breakthrough of the radioactive 3 H, 36 Cl and fluoresceine and Amino-G dyes showed similar behavior. On the other hand, no breakthrough of montmorillonite colloids or 152 Eu occurred. Based on the literature review, the low flow rates used could be the reason for this. Low flow rate (50 μl/min) could affect the colloid mobility strongly which could explain that Eu retained in the fracture. More experiments with higher flow velocities would be required. Different montmorillonite materials showed similar but not exact the same sorption behavior of Eu. The fraction of Eu attached to colloids decreased during the experiments and correspondingly the fraction attached to the granite increased. At the same time, colloids remained stable during the expertiments. This indicates that desorption of Eu from the colloids is taking place in the presence of granite. Also, the effect of different water composition on the stability of colloids was clearly seen on the preparation of colloid suspensions in different water simulants. Even a small increse in the ionic strength of the solution made the especially Ni-montmorillonite colloids instable. Avainsanat Nyckelord Keywords Bentonite, montmorillonite, colloids, final disposal of spent nuclear fuel, Laser-induced breakdown detection Säilytyspaikka Förvaringställe Where deposited E-thesis Muita tietoja Övriga uppgifter Additional information

3 Contents Abbreviation list Introduction Final disposal of spent nuclear fuel Experimental methods for determining radionuclide transport Batch sorption method Column method Electromigration Block-scale experiments In-situ experiments Colloids Terms and definition of colloids Bentonite colloids...17 Structure of clay minerals Relevance to the safety analysis of the final disposal...21 Stability of colloids Mobility of colloids Bentonite colloids in context of safety assessment Colloid Interactions with radionuclides and other pollutants...29 Radionuclide uptake Europium chemistry and use as an analog for trivalent actinides Theoretical background of experimental methods Colloid Detection...35 Laser-induced breakdown detection Photon Correlation Spectroscopy Radioactivity detection...50 Gamma radiation spectroscopy Liquid scintillation Experimental Work Materials...52 Kuru Gray granite block Water simulants

4 Montmorillonite colloids Tracers Methods...61 UV-VIS spectrometry Photon Correlation Spectroscopy Laser induced breakdown detection Radioactivity Measurements MP-OES Experimental settings...67 Block scale experiments Batch experiments Results and discussion Colloid characteristics Eu distribution between the granite, colloids and ground water Block-scale experiments on radionuclide migration Discussion Conclusions References Appendices Appendix 1: Equations for the activity calculations Appendix 2: Composition of Grimsel Groundwater Simulant used within this work (Suorsa) and other composition reported in the literature Appendix 3: Original 152 Eu-solution information provided by the manufacturer. 117 Appendix 5: Quench series for 3 H measured with liquid scintillation detector

5 Abbreviation list AGA Amino-G dye (7-Amino-1,3-naphthalenedisulfonic acid monopotassium salt monohydrate) CCD Charge-Couple-Device GGW Grimsel ground water HLW High Level Waste KBS-3 Kärn Bränsle Säkerhet 3 KIT-INE Karlsruhe Institute of Technology/Institute for Nuclear Waste Disposal LIBD Laser induced breakdown detection PMMA Poly(methyl methacrylate) PCS Photon correlation spectroscopy SNF Spent Nuclear Fuel UH University of Helsinki 3

6 1. Introduction In Finland, the spent nuclear fuel (SNF) from the nuclear power plants will be deposited in crystalline bedrock at the depth of several hundred meters. The safety of waste disposal has to be secured for hundreds of thousands of years and indeed, intensive knowledge has already been gathered to assess the phenomena taking place during this long-term process. An important aspect is the possible groundwater intrusion after a failure of the designed artificial barriers. Colloids have been found to have an impact on the mobility of many otherwise immobile radioactive contaminants. One potential source of the colloids has been assumed to be the bentonite clay used as a release barrier. In contact with dilute groundwater the bentonite barrier could degrade and the colloids be mobilized. This could lead to a potential migration of radionuclides from the deposit to the biosphere in the case of canister failure. The colloids pose a threat to the long-term safety because they could act as a mobile phase and therefore enhance the mobility of e.g. otherwise immobile transuranium elements. In normal conditions, the bedrock would act as an ultimate barrier after all the other barriers would have failed. The retarding effect of bedrock is based on the sorption reactions at the fracture surfaces and the diffusion inside the rock matrix. The mobile colloids could however offer a vector to prevent these mechanisms. The clay colloids also have a small size, having a large surface to mass ratio and therefore sorbing effectively cationic elements. A large amount of data is needed to understand the interactions between the radionuclides, clay colloids and bedrock. The topic culminates in questions like is there a source of colloids, are the colloids stable and mobile, do the radionuclides sorb to them strongly enough to pose a significant threat. All these questions cannot be answered by simple laboratory experiments. In addition, field studies have been conducted and are ongoing at the moment. These in-situ experiments have however some restrictions like not having a full control of the flow conditions. The large scale laboratory experiments done with e.g. granitic blocks offer an intermediate between the two ultimate scales. The objective of this work was to assess the suitability of a granitic block with a natural fracture to the colloid radionuclide interaction experiments. The intention was to gather 4

7 knowledge on the stability of flow paths and to examine the mobility and sorption behaviour of montmorillonite colloids in environmentally relevant conditions. Different sophisticated methods such as photon correlation spectroscopy (PCS) and laser induced breakdown detection (LIBD) were utilized for the bentonite colloid detection. The work for this study was conducted in the Laboratory of Radiochemistry in University of Helsinki (UH) and Karlsruhe Institute of Technology (KIT), Germany. 2. Final disposal of spent nuclear fuel KBS-3 concept In Finland, the spent nuclear fuel from the five reactors of two nuclear power plants, Loviisa and Olkiluoto, will be disposed at a depth of m in the crystalline granite in Olkiluoto. The final disposal is based on the KBS-3 concept developed by Swedish nuclear waste management company SKB. The concept relies on the multi-barrier principle, where the fuel itself, copper canister, bentonite buffer and the surrounding bedrock prevent the migration of radionuclides to the biosphere from the disposal site (Figure 2.1) (SKB, 2010a; Posiva Oy, 2012) Figure 2.1. The Multiple-barrier principle developed by SKB. Picture taken from (Posiva Oy, 2016). The basic principle is that the spent uranium fuel pellets inside zirconium assembly will be placed inside disposal canisters made of copper and cast iron 5

8 insert. The cast iron will offer mechanic protection whereas the copper acts as corrosion cover. The number of these canisters is estimated to be around (Posiva Oy, 2016) The fuel itself is mostly in chemically inert and immobile form as UO2 in the conditions of the disposal site but it is also accompanied by the chemically diverse fission and activation products. These products cause the most of the safety issues related to the final disposal of spent nuclear fuel. An important remark of the behaviour of the spent nuclear fuel is that it is highly dependent on the stability of UO2 which in turn depends considerably on the prevailing redox-conditions. (Bruno and Ewing, 2006) The fuel closed in the disposal canisters will be placed into deposition holes in the disposal tunnel excavated to the bedrock. The holes will be tightened up with buffering bentonite clay material which has a property to swell in the contact with water. This provides good insulation for the canister from the surrounding water and also protects it from the movements of the bedrock. The bentonite has also high capacity to sorb cationic species. This makes it also an effective barrier preventing the mobilization of radionuclide cations of the spent fuel. Finally, the bedrock acts as the final barrier which retard the mobility of the radionuclides owing to sorption reactions on rock surfaces and diffusion inside the rock. The composition of the spent fuel The uranium fuel, usually UO2 or U, is not very radioactive. In fact, it is less radioactive than natural uranium ore since the chemical purification steps have removed all the radioactive daughters of uranium which contribute most of the radioactivity in the ore. (Bruno and Ewing, 2006) The composition of the spent fuel consists approximately 95 % of the original material, UO2. The exact fraction is depending on the burn-up of the reactor. Relatively small fraction of the initial material has transformed to fission products, transuranium elements and activation products. Although the fraction of transformed uranium is relatively small, the activity has grown enormously, by a factor of a million. This is because the most of fission and activation products have much shorter half-lives in comparison with uranium and therefore emit more intensive radiation. It takes several hundred thousand years until the radioactivity has decayed to the level 6

9 of natural uranium ore. (Bruno and Ewing, 2006; Ewing, 2015) The activity of the different species can be compared to original uranium ore (Figure 2.2). The fission and activation products are to occur in a highly diverse range of different chemical and physical species. For example, there will be fission-product gases like Xe and Kr, metallic fission products like Mo, Tc, Ru, Rh and Pd, fission products like Rb, Cs, Ba, Zr which occur as oxide precipitates and fission products as Sr, Zr and rare earth elements which occur in the form of solid solutions. There will be also transuranium elements like Pu, Np, Am and Cm within the UO2 matrix. (Bruno and Ewing, 2006; Ewing, 2015) As the fission and activation products are radioactive, the composition of the spent nuclear fuel will change in the course of time. In the beginning, most of the radioactivity is caused by the short-lived fission products and the activity decreases quite quickly. After 10,000 years, just 0.01 % of the activity is left compared to the activity one month after the removal from the reactor. (Hedin, 1997) Evidently, the extent of the potential risk is dependent on both the half-life of the radionuclide, its radiation properties and the chemical characteristic. It is not only the radioactivity but also the chemistry of a particular element which contributes to the biological hazard. If a highly radioactive radionuclide is retarded because of its chemical nature, it does not pose a threat for biosphere. This retardation can of course be compromised because of changing conditions. The evaluation of these changes is an essential of the safety analysis. The predominant radionuclides contributing to the penetrating gamma-radiation are 137 Cs and 90 Sr, which both have a half-life of about 30 years. The half-lives are still so short that these are not important regarding the long-term safety of the final disposal of the spent fuel, although they have to be taken into account in the operative phase. In the long term, the most remarkable radionuclides are long-lived fission products and transuranium elements which may be mobile and hence migrate from the disposal site. Some such examples are 99 Tc, 129 I, 79 Se, 135 Cs, 239 Pu, 240 Pu, 241 Pu and the daughter of the latter, 241 Am, as well as even more long-lived 243 Am. After a few hundred years 137 Cs and 90 Sr have decayed, 241 Pu and 241 Am are the nuclides, which contribute most to the radiotoxicity. (Ewing, 2015) 7

10 Figure 2.2. Radioactivity of high level nuclear waste (HLW) in one ton of spent fuel. (IAEA, 1992) Possible scenarios It takes hundreds of thousands of years until the radioactivity of the spent nuclear fuel descends to the level of natural uranium ore, accounting for the activity of the daughters. The timescale of the final disposal being extensive, many different scenarios must be taken into account. For example, in the evalution of the conditions in fuel repository done by Pastina and Hellä (2006), the authors take into consideration two glacial periods. Still the performance of the repository is estimated to perform well. For example, most of the copper canisters are evaluated to be still acting as a barrier after 125,000 years. After the radiation has decayed to the level of natural uranium ore, the disposal concept has fulfilled its requirements. 8

11 The main effects of the glacial melt water intrusion in the context of colloids would be the dilution of the groundwater, which could enhance the stability and mobility of the colloids. In addition, the higher concentration of dissolved oxygen would affect enormously the mobility of uranium found in the fuel. In the case of oxygen intrusion, the retarding fuel matrix would be lost. The break of copper canister at the same time would lead to the leakage of radioactive material from the disposal site. On the other hand, the fuel will be surrounded by cast iron of which oxidation will consume most of the intruding oxygen. (SKB, 2010b) After all it is about the physico-chemical conditions prevailing at the depth of the final disposal which are important regarding the performance of multiplebarrier system. The changes in the conditions will often affect different barriers although the mechanisms are different. For example, the glacial water would dilute the water enhancing the bentonite buffer s chemical erosion and in the same time oxygen content in the fresh glacial melt water will accelerate the corrosion rate of the copper canister (Pastina and Hellä, 2006). 3. Experimental methods for determining radionuclide transport The behaviour of the radionuclides and the interaction with the retention barriers can be examined with many different experimental arrangements. Not only the manner of approach but also the scale of the experiment has to be adjusted to best suit the point of interest. The physical and chemical behaviour can be examined with small and easily reproducible laboratory batch experiments but also with large scale laboratory and even more complex in-situ experiments. The data from both and everything between is still essential to understand the real phenomena taking place in the final disposal conditions. On the other hand, deep understanding of the phenomena is needed to perform the safety analysis for the final disposal of spent nuclear fuel. 9

12 The simplified methods are often needed to gain the basic knowledge of the interactions of different radionuclides with different barrier materials. It may be challenging to transfer the knowledge gained from these kinds of experiments to natural, complex system. For example, the results from contact experiments of radionuclides with different purified mineral phases do not fully represent the behaviour found from the natural material. Also, it is a question how does the static and dynamic experiments correlate with each other. (Hölttä, et al, 2001) Another aspect is the use natural analogues, i.e. observing the nature (McKinley and Scholits, 1993), but these are not discussed further within this thesis and are covered comprehensively by Miller et al. (2000). 10

13 3.1. Batch sorption method All different sorption processes grouped together are characterised by the mass-based distribution coefficient Kd (see Equation 3.1). C solid K d Equation 3.1 Csolution where Csolid = concentration of nuclides per solid mass (mol kg -1 ) Csolution = concentration of nuclides in the solution (mol m -3 ) The batch sorption experiments are simple experiments, where the sorbent and the solution are mixed together in a vial. The two phases are separated after a convenient period of time by e.g. filtering or centrifuging. (Loebenstein, 1962) Then the distribution of a contaminant or another kind of an analyte is examined. The batch sorption experiments offer a versatile, easy and a practically reproduced arrangement to study interaction of radionuclides in simplified conditions. (Kautenburger and Beck, 2010) The authors note the batch experiments as a state of the art technique offering a good way to compare different results. Many different phenomena associated to the performance assessment have been studied with the batch experiments. For example, radionuclide sorption to rock (Jin, et al, 2014; Tachi, et al, 2015) (Puukko and Hakanen, 2004; Huittinen, et al, 2010; Schnurr, et al, 2015a; Elo, et al, 2017) or (Huber, et al, 2011; Huber, et al, 2015) can be investigated by the batch experiments. In addition, the effect of microbes to the radionuclide retention have been studied with batch experiments (Lusa, et al, 2015a; Lusa, et al, 2015b; Lusa, et al, 2015c). The limitations of the batch scale are related to the extending of the results to the larger, natural scale. For example, the mineralogical composition of the rock plays a large role as well as the specific surface area. With freshly crushed rocks new surfaces are created which affect the distribution of radionuclides between the solid and liquid. (Missana, et al, 2006; Tachi, et al, 2015) The static batch experiments offer a good way to gain 11

14 database of KD-values but larger scale experiments are needed to fully understand the behaviour of radionuclides in the natural system. (Missana, et al, 2006) 3.2. Column method Dynamic flow-through experiment using crushed rock or fracture column is a better approach than static batch experiments for determining the parameters affecting the radionuclide transport. Column experiments allow the realistic solution to solid ratio for the calculation of Kd-values. Column method has been used in many experiments related to radionuclide transport in rock material. For instance, the matrix diffusion (Hölttä, et al, 1996; Hölttä, et al, 1997), colloidradionuclide interaction (Dittrich, et al, 2015; Elo, et al, 2017) and rockradionuclide retention (Dittrich and Reimus, 2016). The disadvantages of the column method include the time needed to perform the experiment, especially in the case of strongly sorbing radionuclides. Compared to batch experiments, the column experiments offer a more convenient arrangement, where also the water-flow is present Electromigration The diffusion of radionuclides through a solid rock is a slow process, which restricts the use of direct experiments. However, the electric potential can be utilized to speed up the process by inducing electromigration and electroosmotic flow through the rock. This is achieved by using two electrodes on the both sides of the sample and the electrical potential gradient drives the charged radionuclides through the sample. For example, such a method have been utilized to study the diffusion of radionuclides in clay (Maes, et al, 1999) and the sorption of Cs + to different rock materials (André, et al, 2009; Puukko, 2014). The theory of electromigration is comprehensively covered by (André, et al, 2009). 12

15 3.4. Block-scale experiments Upscaling from the column experiments, the block scale is a direct approach to the natural conditions but still having the versatility and being relatively simple and cost-effective arrangement compared with the field experiments. Furthermore, the flow field characterization is difficult or even impossible in the in-situ conditions, which requires quite much of approximation and might involve much of conjecture. (Vandergraaf, 1995) The block scale experiments fill the gap between the small laboratory and large-scale in-situ experiments. In the block scale experiments the studies are carried out in a block having water conducting fracture. Within the studies related to radionuclide migration, different phenomena have been investigated in the block scale. Matrix diffusion (Hölttä and Hakanen, 2002; Hölttä, et al, 2004; Hölttä, et al, 2008), the general radionuclide migration (Vandergraaf, et al, 1996; Park, et al, 1997) and the effect of nanoparticles to the migration (Vilks and Bachinski, 1996; Vilks and Baik, 2001; Vilks, et al, 2008) have been studied with granitic blocks. The advantage of the block scale is that much better control of the system is usually achieved compared with the field studies. The better control of ph, pressure and chemical conditions is beneficial. This helps especially when studying matrix diffusion where the low flow rates are needed for the phenomenon to be visible. (Hölttä and Hakanen, 2002) Especially the work of Peter Vilks and co-authors with the granitic block concerning the colloidal transport of radionuclides (Vilks and Bachinski, 1996; Vilks, et al, 1997; Vilks and Baik, 2001; Vilks, et al, 2008) is interesting in the context of this thesis. For example, the effect of groundwater flow rate was crucial to the colloid stability. Vilks and Bachinski(Vilks and Bachinski, 1996) conducted experiments with a granitic block in a dipole system where the injection in the horizontal fracture was done from one borehole and collection was from another. They found out a strong correlation between the flow velocity and the colloid recovery. In their later work Vilks and Baik (2001) found out 13

16 significant differences in the breakthrough curves of different elements with and without colloids. For example, divalent Sr 2+ came at the same time with the conservative tracer, Br -, when associated with colloids but remarkably retarded without. With trivalent Am, the effect was even more remarkable, as no breakthrough at all was observed without colloids. With colloids, the Am breakthrough occurred at the same time with the conservative tracer In-situ experiments The natural environment is representing best the conditions prevailing in the final disposal sites. On the other hand, it might be difficult to avoid altering the conditions, when building up the research facility. Anyway, the studies in the field scale are important giving the knowledge of the behaviour in the natural site. Many colloid-related studies have been done in relevant locations for the final disposal of spent nuclear fuel, for example in Finland (Laaksoharju, et al, 1993; Vuorinen and Hirvonen, 2005) and in Sweden (Hauser, et al, 2005). Some in-situ studies have been also conducted more specifically related to the colloid mobility. Such work has been done in Grimsel Test Site in Switzerland within the Colloid and Radionuclide Retardation Experiment (CRR project), where bentonite colloid assisted radionuclide transport has been studied within a rock fracture. (Fleming, Crerar 1982, Stumpf, Stumpf et al. 2008, Allard, Beall 1979)(Hauser, et al, 2002; Möri, et al, 2003a; Geckeis, et al, 2004) At present, the Colloid Formation and Migration (CFM) is underway to study colloid generation and migration at the same site. The site is characterised by natural conditions representing well the conditions relevant for the long-term safety of final disposal in crystalline bedrock. (Nagra, 2017) 14

17 4. Colloids 4.1. Terms and definition of colloids A term colloid means a phase (gas, solid, liquid), which is evenly distributed in another phase with a different composition (IUPAC, 1997). For example, solid particle in gaseous phase is an example of a colloidal mixture. Such is called smoke in everyday life. Another example is milk, where e.g. animal fat is dispersed in water. In the context of radionuclide mobilization, one is dealing mostly with solid particles suspended in water which can be of organic or inorganic nature. In Figure 4.1 two dispersion are presented, the stable and unstable to point out the crucial difference between the two systems. Figure 4.1. The difference between stable and unstable colloidal dispersion. In the left vial the particles are evenly dispersed because of their internal movement, the Brownian motion, is stronger than the gravity and the interactions with another particles. In the right vial the particles are agglomerating, i.e. smaller particles are attached to each other forming larger clusters, which are going through sedimentation because of the gravity. The stability of colloidal dispersion is dependent on the size of the particles, or colloids. If the particles are too large, they are likely to sediment because the gravitational force outcomes the Brownian motion. The Brownian motion describes the particles random movement because of continuous interactions with the molecules in the liquid. Even though the particles would initially be 15

18 small enough to stay suspended in media, they might agglomerate as can be seen in right vial in Figure 4.1. Therefore, it is also the prevailing conditions, which strongly affect the stability of the colloids. The most important parameters affecting the colloid stability are ph, ionic strength and temperature. The ph affects strongly the surface charge of the particles. The isoelectric point (pi) is a ph value when certain particle s net electric charge is zero (IUPAC, 1997). Particles having the same net charge repel each other. At isoelectric point these repelling electro-static forces are at their smallest meaning that there are less forces to prevent particles interacting with each other. It is not unambiguous to say what size range of particles is stable since it is strongly dependent on the composition of particles. The density of particles is essential in this context of the upper size limit. (Geckeis, et al, 2011) The size limit for colloidal properties is therefore related to the material of the particles (Hochella Jr., et al, 2008). Although, it seems that the relevant colloids in the context of contaminant transport have at least one dimension which is less than 100 nm. For example, Novikov et al. (2006) found out that % of Pu was mobilized by colloid with diameter of 1-15 nm. Inorganic colloids can be divided into two classes: eigen- and pseudo-colloids. The former is sometimes also called intrinsic- or true-colloids. (Kudo, 2001) They are polymerized clusters of an element. For example thorium (Bitea, et al, 2003; Bundschuh, et al, 2000) and plutonium (Knopp, et al, 1999) colloids are proved to exist and to remarkably enhance the solubility of these actinides at their tetravalent state. (Geckeis, et al, 2011) Pseudo-colloids act as carriers but are not composed of the radioactive materials itself. The radionuclides or other contaminants can be attached to these carrier colloids, which makes the otherwise immobile species mobile. The carrier colloids can be silica, clay minerals, calcite or metal-oxyhydroxides. Especially the clay minerals are interesting in the pseudocolloid point of view. The nanoparticles of mineral material can be found widely from the environment. They are dispersed in the oceans, fresh waters and atmosphere. (Hochella Jr., et al, 2008) Another class is organic colloids like microbes and fulvic or humic acids (Geckeis, et al, 2011). The two latter are the ultimate form of the degrading organic material and these 16

19 are present almost everywhere in the nature, especially in the surface waters. The mechanism they could affect radionuclide migration is by stabilizing the inorganic colloids by making those more stable. This is supposed to be because of creating of organic coatings to inorganic colloids. (Geckeis, et al, 2011) Regarding the final disposal of spent nuclear fuel, many kinds of different colloids may contribute to the performance of the disposal concept. These nanoparticles can be organic or inorganic and in the size range from 1 to 1000 nm. It is obvious that they have to remain suspended in water.(geckeis, et al, 2011) The colloid can be naturally occurring at the disposal site but also colloids from the buffer materials are likely to exist with the time frame of the final disposal (Schäfer, et al, 2012) Bentonite colloids Structure of clay minerals One of the most important colloid sources in the context of final disposal of spent nuclear fuel are clay minerals, especially bentonite which will be extensively used as a buffer material. The bentonite consists of two different sheets: the tetrahedral and octahedral. The tetrahedral sheet is made of silica tetrahedra, where a silicon atom is having a bond with four oxygen atoms. The octahedral sheet consists of one aluminium atom attached to six oxygen atoms. These sheets are connected when the Si and Al are sharing an oxygen atom between them. This way there is a three-sheet structure called a layer. The sheets continue in the lateral direction and many of these sheets are stacked together forming a crystal lattice. The bonds keeping the sheet together are covalent by nature. On the contrary, the different sheets are just staying together because of much weaker Van der Waals and electrostatic forces. (Luckham and Rossi, 1999) 17

20 Figure 4.2. The structure of montmorillonite. The three sheets, which form a layer can be seen on the bottom of the picture. Above another layer is partially shown. The picture taken from Luckham and Rossi (1999). The cations in the clay structure are sometimes substituted with similar sized cations, but with different charge. For example, Al 3+ or Fe 3+ can substitute Si 4+ in tetrahedral sheet. Similarly, the Al 3+ in octahedral sheet can be replaced by Mg 2+ or Fe 2+. (Luckham and Rossi, 1999; Pastina and Hellä, 2006) All these substitutions give rise to charge deficiency. Furthermore, this results in negative charge on the clay layer, which is compensated by exchangeable cations, e.g. H +, Na +, Ca 2+. These cations can be changed with the ones in the bulk solution. (Luckham and Rossi, 1999) Different clay minerals have a layer structure. Yet, there are big differences between the different minerals. Illites and kaolinites layers are located so close to each other that no water can penetrate between them. The reasons for this are high layer charge (illite) and strong hydrogen bonds (kaolinite), which result in strong interlayer bonding. Smectites have a 2:1 structural unit where the octahedral sheet has tetrahedral units on the both faces. (Luckham and Rossi, 1999; Pastina and Hellä, 2006) This means that the tetrahedral sheets of layers are facing each other and the oxygen atoms of tetrahedral sheets are located opposite to each other causing repulsion and therefore weak bonding. In addition, the substituted atoms can also contribute to the repulsion, even strongly in some situations. The weak bonding of two layers results in increasing 18

21 space between them. This enables the water penetration to the empty space, which is unique to smectite clays. (Luckham and Rossi, 1999) The most important of the smectite group in the context of the final disposal of the spent nuclear fuel is montmorillonite because the bentonite clay used as buffer mostly consisting of it. One third of cations in the montmorillonite s octahedral sheets are displaced by divalent cations in this mineral. Basically Al 3+ is replaced by Mg 2+. The charge is then balanced by exchangeable cations like Na +, Ca 2+ and Mg 2+. The class of montmorillonite is often remarked by the major exchangeable cation. Two of most familiar in the context of final disposal of the spent nuclear fuel are Na- and Ca-montmorillonite. Pastina and Hellä (2006) consider both as reference materials. MX-80 is Na-montmorillonite from Wyoming, USA and Deponit CA-N is Ca-montmorillonite from Greek. The former is the material planned to use in the repository. The bentonite s ability of retain cations is based on its structure. The majority of cations are bind in the interlayer space. Extent of the interlayer-space is dependent on the chemical characteristics of these cations. If they are monovalent cations like Na + or Li +, the tetrahedral sheets are repelling each other stronger. If there are multivalent cations like Ca 2+, the repulsion is not as strong and the sheets are located with a narrower space between them. In a same way the concentration of cations affect the space between the two layers but it is all about the ionic strength. In other words, same concentration of divalent cations is affecting much more strongly than monovalent. Essentially, the surface charge is same with high and low ionic strength. The difference comes out from the size of the diffusion layer between the two sheets. (Luckham and Rossi, 1999) In Figure 4.3 the ion concentration is presented as a function of the distance from the particle interface. The graph tells that the diffusion layer has a remarkable effect on the cation attraction. Not only the ions just next to the interface are electrostatically interacting with the solid surface, but also the ions further away. In summary, this means that with a higher ionic strength this diffusion layer is diminished and there is no place for the cations. In the extreme situation, the two tetrahedral layers are located next to each other so, that there is just one 19

22 diffusion layer between the two interfaces. This means that there is much less space for the exchangeable cations to stay. (Luckham and Rossi, 1999) Figure 4.3. The concentrations of cations (n+) and anions (n-) as a function of distance from the surface. The cation concentration is going slowly down. This means that there is still some electrostatic attraction even behind the ions just next to the interface. Picture taken from Luckham and Rossi (1999). The interactions, which cause particles to aggregate, are happening at different interfaces of the particles. Since bentonite is formed from the layer structure, there are two kinds of surfaces: the faces or basal planes and the edges. The faces have the large surface area, while the edges have much smaller. In nature, the particle interactions can be face-to-face, edge-to-face or edge-to-edge. There is a big difference between these interactions. The face-to-face contact best describes the aggregation. Then the two tetrahedral sheets are located very near each other and there is no diffusion layer for the cations to be. On the other hand, edge-to-edge and edge-to-face interactions could lead to formation of bigger molecules, which would basically still have the active diffusion layers. There are quite many models to describe the clay particle associations with each other, but it is not an easy task to conclude theoretically. (Luckham and Rossi, 1999) An important factor related to nanoparticles is zeta potential, which is related to the surface charge of the particle. Larger charges on the surface, the bigger the affected water molecule layer is formed around it. The zeta potential is derived 20

23 from the particle velocity in an electrical field. The subject is treated in more detail in Photon Correlation Spectroscopy section of chapter 5.1 Colloid Detection in this thesis Relevance to the safety analysis of the final disposal Classically, the concept of contaminant transport has included only two phases: the stationary sediment or bedrock and the mobile groundwater (Geckeis, et al, 2011). The concept ignores the effect of mobile nanoparticles, i.e. colloids, and could lead to a remarkable underestimation of the mobility of the contaminants. Obviously strong enough interaction between the colloids and contaminants is essential to make mobilization possible.(kersting, et al, 1999) On the other hand, Geckeis, et al. (2011) mention that the colloids could offer new retention mechanisms, especially for strongly sorbing radionuclides. On the other hand, there are many mechanisms, which could retard the colloids. Attachment to surfaces of surrounding rock matrix, agglomeration, sedimentation and filtration in dense material could make the colloids immobile. This would also affect the mobility of sorbing radionuclides.(geckeis, et al, 2011) Also the desorption is having an important role in the mobility. The radionuclide sorption must be irreversible to have an effect in a long-term safety (Miller, et al, 2000),, which is described in detail later in this study. The effect of colloids is often described with a step model that includes many of the previous aspects (Miller, et al, 2000; Möri, et al, 2003b; Schäfer, et al, 2012). In Figure 4.4 these terms are put in a schematic picture. Colloids must be stable to affect the safety final disposal. They have to be also mobile and stable, as well as they must interact with radionuclides to act as carriers. Finally, this uptake must be irreversible, which means that the radionuclides are not losing their interaction with the mobile colloids. If any of these terms is not fulfilled, the colloids are not significant in the prevailing conditions. (Miller, et al, 2000) On 21

24 the other hand it should be born in mind that the conditions can change during the long-term disposal. Figure 4.4. Colloid ladder that demonstrates the fact that all the terms in the centre has to be fulfilled to make colloids significant to the safe final disposal of spent nuclear fuel. Picture taken from Miller, et al. (2000). All these terms or questions are important from a scientific point of view but on the other hand it is essential to assess the priority of the different aspects. If there would be no colloid present, it would be waste of resources to research such a topic in the context of the safety assessment of final disposal of nuclear fuel. On the contrary, all of these terms are more or less dependent on the prevailing conditions. That is why all the questions are important to investigate and especially vital is to know their relations to different physico-chemical conditions. 22

25 In the crystalline rock and in conditions geochemically and hydraulically near to equilibrium, the inorganic nanoparticles are mostly limited to concentrations of <1 mg/l and their stability is also dependent on the ionic strength of solution. The concentration of the alkaline earth elements has a great effect to that. In geochemically and hydraulically disturbed systems, also higher concentrations are possible. (Geckeis, et al, 2011) In more saline solutions the concentrations seem to be much smaller, e.g. Laaksoharju and Wold (2005) state that the colloid content in Äspö is less than 0.3 mg/l and in repository level less than 0.05 mg/l. Wold (2010) determined the colloid concentrations in Laxermar and Forsmark and found out that they are mostly less than 60 g/l. Degueldre, et al. (1996) measured the natural colloid concentration in Grimsel to be around 0.1 g/l, which seems quite low indicating that the ionic strength is also very low (1.2 mmol/l) and therefore the conditions should be favourable for the colloid stability. On the contrary, the prevailing conditions in Olkiluoto do not favour the colloid formation and stability (Pastina and Hellä, 2006). However, the higher colloid concentrations can be take place in geochemically or hydraulically disturbed conditions. Changes in temperature, water flow and chemical parameters can lead to higher colloid concentrations. (Degueldre, et al, 2000) It is evident that some or even all of these parameters are going to change during the final disposal of the spent fuel. Stability of colloids The theoretical background for colloid stability is based on DLVO-theory developed by Derjaguin, Landau, Wervey and Overbeek abbreviation coming from the first letters of the inventors. According to the theory, the interactions between colloidal particles are determined by the attractive Van der Waals forces and the repulsive surface charges. (Alonso, et al, 2006) The stability of colloids is depending on several variables in the natural environment. The stability is strongly dependent on their possible agglomeration. This is mostly controlled by the colloid surface charge which depends on the solution composition like ph and ionic strength. (Geckeis, et al, 2011) The effect of ph is direct, since the concentration of hydronium ions determines how much of the 23

26 mineral surface s -OH groups are protonated or not. The effect of ionic strength is similar but not as straightforward. The ions, especially the divalent cations, are screening the surface charge of mineral surfaces and thus reducing the repulsion.(geckeis, et al, 2011) In natural waters the most important elements is mainly Ca 2+ but also Mg 2+. For montmorillonite colloids the stability borderline lies for Ca 2+ at mol/l, the limit being almost independent of ph. (Seher, 2010; Schäfer, et al, 2012) This border is called critical coagulation concentration (Ca-CCC), which means that the colloids are stable below this concentration. On the other hand the similar parameter for Na + (Na-CCC) is mol/l at ph = 6 but mol/l at ph = 9.(Seher, 2010; Schäfer, et al, 2012) Essentially, there can be seen a remarkable change over the natural ph-range. In principle, the CCC is much higher for monovalent cation like Na +. Similar results has been obtained by Missana, et al. (2003). The studies of bentonite colloid stability as the function of ionic itrength and ph were performed with FEBEX bentonite. The stability of colloids was found to be quite constant in the ph range between 8 and 11. Below that the stability was significantly lower. (Missana, et al, 2003) Lahtinen, et al. (2010) investigated the formation of colloids from bentonite and granite material. They found out that particle concentrations were much lower in saline OLSO water than in Allard water, a reference ground water for low-salinity waters, at all phs. Similar findings were discovered previously in the work made by Hölttä, et al. (2009). The silica colloids were found to be stable in Allard and in purified MilliQ water, but not in OLSO. For organic colloids, such a stability criterion as CCC is difficult to obtain. They seem to remain stable in quite broad range of ph and ionic strength conditions.(geckeis, et al, 2011) The interesting point of view is the possible ability of organic colloids to stabilize the inorganic colloids by forming coatings. Wang, et al. (2004) discovered that the combined effect of γ-alumina and humic acid contributed together to the sorption behaviour of Cm 3+. Also Choppin (2003) points out that organic macromolecules, like humics, can enhance strongly the sorption of hydrolyzed species to colloids. 24

27 The temperature does not have such a remarkable effect on the colloid stability as the ph and the ionic strength (Ryan and Elimelech, 1996; Hölttä, et al, 2009). Although, García-García, et al. (2006) conducted experiments with bentonite and latex colloids in which they found out that the bentonite colloids could be even more stable in higher temperatures. The opposite results were found with amidine latex particles. In further experiments (García-García, et al, 2009) concluded that the in the natural ph range 7-8, the temperature could stabilize the colloids strongly. According to Choppin (2003), the temperature could affect the diffusion rates of colloids but also the speciation of radionuclides, for example plutonium. Mobility of colloids The main contributors to colloid mobility are the stability and the retardation processes on the flow-path. The stability was discussed in the previous section. If the colloids are not stable they evidently do not move far enough to affect the performance assessment. The retardation processes taking place in the rock matrix are, for example attachment on to the surfaces, matrix diffusion and sizediscrimination. These all could be combined under the term filtration.(alonso, et al, 2007a; Schäfer, et al, 2012). A good summary of the retardation processes is presented in a figure by Kretzschmar and Schäfer (2005) (see Figure 4.5.) Additionally, the filtration of colloids in the bentonite buffer is likely to take place at the repository site. (Alonso, et al, 2011) 25

28 Figure 4.5. The colloid mobility in saturated (a) and unsaturated (b) media. The retention processes affecting the colloid mobility: 1. Colloid formation, 2. Colloid aggregation, 3. Gravitational immobilization, 4. Immobilization of single particles or particle clusters, 5. immobilisation to the surfaces, 6. Colloid transport with flow. For unsaturated system there are also additional retention mechanisms: 7. Film straining and 8. Immobilization in gas-liquid interphase.(schäfer, et al, 2012) Figure taken from Kretzschmar and Schäfer (2005). It seems that the colloid mobility and stability is affected strongly by the flow velocity under advection-controlled systems. The possible colloid transport is then likely to prevail in conductive fractures.(alonso, et al, 2006) Also the roughness of the fracture surfaces influences the colloid mobility and stability in the experiments. These both contribute strongly to the colloid retention on the rock surfaces. (Schäfer, et al, 2012) The size of colloids is another variable affecting the retention (Alonso, et al, 2006). The colloid attachment on to the surfaces depends on the ph and ionic strength as does the colloid stability: the electrostatic attraction enables the attachment of the colloids to fracture surfaces (Schäfer, et al, 2012). This is logical since the process is a solid-solid interaction similar to the interactions between the colloids, although another solid is in the form of colloids and another as bulk material. In the case of agglomeration both solids are colloidal. Under the conditions favourable for the attachment the surface charges of the rock and 26

29 colloids are opposite. On the contrary, when the charges have the same sign the conditions do not favour to attachment. It is still remarkable that the deposition of colloids on to rock surface can be observed also under the conditions unfavourable for attachment. Alonso, et al. (2009) conducted experiments with gold nanoparticles and granite. The experiments were done under the favourable (ph 5) and the unfavourable (ph 9.5) conditions for the retention. The gold particles were used as an analog for bentonite particles since they are easier to examine inside granite by μpixe. The method is not suitable for bentonite colloids because of the high Al and Si content.(alonso, et al, 2009) The outcome was as expected in a way: under favourable conditions the attachment was mostly higher compared with the unfavourable conditions. The extent of attachment was also dependent on the particle size: the larger particles were more strongly attracted. Nevertheless, deposition on the mineral surfaces occurred also under unfavourable conditions. The extent of attachment reached non-negligible values in the unfavourable conditions, remarkably without a clear size dependency. Alonso et al. explain that the reason for the behaviour is that the retention is strongly affected by the small-scale chemical effects as well as the heterogeneity of the rock surfaces. In similar studies Albarran, et al. (2013) found that a similar behaviour takes place with smectite colloids also. In column experiments the colloid recovery, which means how much of initially injected colloids were detected after the migration through the column, was dependent on the flow rate. The smectite recovery was over 30 % with the slowest flow velocities (16 μl/min) used. With higher velocities (50 μl/min) recovery was 100 %, meaning that no colloid retention occurred. The gold and latex particles were also investigated and the decrease in the recovery was greater with latex and gold comparing with smectite which the authors explained by the different particle geometry. The gold and latex particles are spherical but smectite particles are platelets. The geometry could strongly affect the interaction of colloids with the rock matrix in conditions unfavourable to the colloid retention since with the face-to-surface association the repulsion could be stronger because of larger contact area. This could explain the difference between different particles. Alonso, et al. (2006) 27

30 state that although granite has a negative surface charge in natural ph-range, some of the minerals in it can still have a positive charge. This could be the reason for colloid deposition even in the unfavourable conditions. The classical model seems to fail describing the colloid deposition under the unfavourable conditions. In addition to retention on the surfaces, the matrix diffusion is a process, which may slow down the migration processes of contaminants. In the case of colloids, the effect should still be minor. (Schäfer, et al, 2012) Alonso, et al. (2007b) researched the effect of matrix diffusion and found out that nm gold particles were diffused in to the rock matrix. The rate of diffusion was influenced by the particle size: decreasing size lead to increasing diffusion rate and no diffusion was observed for the 250 nm particles. In diffusion-controlled systems, for example in clay deposit, there is no advection to affect the mobility of colloids. As the definition suggests, the mobilization is only because of diffusion. The colloid concentrations in such systems are expected to be much lower compared with advection-controlled systems.(schäfer, et al, 2012) The colloid mobility is different between saturated and unsaturated medias. The mechanism of migration differs between the systems. Altogether it is not an easy task to understand because of complexity of the system. In unsaturated media, the dynamic air-water interfaces complicate the prediction of the colloid behaviour. For example, film straining or retention on the air-water surface could be possible.(schäfer, et al, 2012) Bentonite colloids in context of safety assessment At the point when all the man-made barriers have failed, the surrounding bedrock acts as the ultimate barrier. Then solubility of radionuclides, sorption reactions with the mineral surfaces and dilution by groundwater are in 28

31 responsibility of the mobility of the radionuclides at the site. (Ewing, 2015) The bentonite colloids could have a remarkable effect to the mobility of radionuclides in suitable conditions as discussed earlier in this section. The question is, will there be such conditions and for a sufficient long time to make the colloidtransport relevant in the large time-scale. At the same time, the reversibility still remains an open question. In the both natural and the repository conditions, the most important factor regarding the colloid stability is the concentration of divalent cations, Ca 2+ but also Mg 2+. Pastina and Hellä (2006) state that it is challenging to predict the colloid formation at specific site like in Olkiluoto. Presently prevailing conditions are not favouring the bentonite colloid stability since the ionic strength of the groundwater is way higher than the critical value (10-20 g/l of total dissolved solids). Yet, the salinity may change because of the glacial waters. That means that it is not possible to rule out the colloid transport. More information is needed on the reversibility processes for example Colloid Interactions with radionuclides and other pollutants Radionuclide uptake Focusing to the carrier-colloid or pseudo-colloids, the essential factor for relevance of colloids to safety assessment is their ability to adsorb radionuclides. The important term is distribution coefficient, Kd. Essentially, it tells the distribution ratio between solid and liquid phase (see Equation 3.1) (Alonso, et al, 2006). Important factor is the reversibility or irreversibility of the sorption. The scenario is that the contaminants, like radionuclides, could be transported with the mobile colloids to the living nature. If the sorption reaction is reversible, it means that the radionuclide is first attached to the colloid surface by some efficiency. After some time, it is although released as the competition between the colloid and bulk surface takes place. If this kind of sorption occurs, the radionuclide transport via colloids is not significant since there is around

32 meters of bedrock between the repository and the earth's surface. In the contrary if the sorption is irreversible, meaning respectively that the contaminants are not released from the surface of the colloid, the contaminant transport can considerably affect the performance of the repository. The schematic picture of the bentonite-radionuclide interaction is presented in Figure 4.6. Figure 4.6. Schematic picture of the mechanisms affecting the colloid-mediated radionuclide transport. The radionuclide uptake and its reversibility are presented in the numbers 5 and 6. Picture taken from Alonso, et al. (2006). The factors affecting the radionuclide speciation affect directly the interactions with colloids. For example, redox conditions and available ligands are important regarding the extent of radionuclide uptake. There are also models developed for this kind of calculations.(alonso and Degueldre, 2003) There are two main surface complexes the actinides and their analogs, e.g. Eu 3+ for Am 3+, can form with solid surfaces: inner and outer-sphere complexes. Geckeis, et al, 2013) The outer-sphere sorption is a process where the cation s first hydration sphere is not disturbed. It can also be described as cation exchange. (Bradbury and Baeyens, 2005; Geckeis, et al, 2013) In the clay 30

33 minerals the ion exchange can occur at the permanent negative charges or in the interlayer region. (Huittinen, et al, 2010) The mechanism of interaction here with the negatively charged mineral is purely electrostatic. This is the prevailing sorption mechanism of actinides in trivalent and pentavalent state to clay minerals at low ph and low ionic strength. Actinides on tetravalent state are so strongly hydrolysed that the electrostatic interactions do not have a noticeable effect to the sorption. The outer-sphere complexation can be regarded almost ph-independent, especially in the low ph. It is still quite dependent on the ionic strength of the solution, because the competing ions conquering the sorption sites. (Geckeis, et al, 2013) On the other hand, the inner-sphere surface complex requires the change in the hydration sphere. One or more of the water molecules are replaced by the surface s oxygen atom. This means that a chemical bond has been formed between the surface and the atom like the actinide in the solution. This makes the sorption stronger since this chemical bond is much stronger than the purely electrostatic bond in the case of inner-sphere complex. This is also affecting the desorption behaviour since the stronger bond is not so easily cleaved. In contrary to inner-sphere complexation, the outer-sphere complexation is strongly dependent on ph, since it is the main parameter affecting the protonation and deprotonation of the surfaces. Also, the ionic strength and solution s chemical composition have an effect. The similar trend can be seen as with the actinide s tendency to hydrolyze: the effective charge of sorbing species is strongly involved in the sorption also. Essentially, this means that the sorption is stronger with a higher effective charge. Another important parameter is the presence of complexing agents like carbonate. (Geckeis, et al, 2013) Bradbury and Baeyens (2005) investigated the sorption competition on montmorillonite with different cations. As a result, they propose that there are different kinds of sorption sites in the surfaces: the weak and strong sites. This was concluded from the fact that just the metals with similar valence seemed to affect each other s sorption. For example, Am(III) and Eu(III) have been found to compete with each other but Th(IV) and U(VI) not (Geckeis, et al, 2013). 31

34 The incorporation processes of sorbed actinides are important since they could affect strongly to their immobilization. There could be different kinds of mechanisms, which would result in the absorption of contaminants into the solid phase. The mechanisms proposed could be mineral dissolution, alteration and the formation of secondary phases.(geckeis, et al, 2013) For example results indicating incorporation of actinides into the solid matrix have been obtained for silica (Chung, et al, 1998), calcite (Stumpf and Fanghänel, 2002) and gibbsite (Huittinen, et al, 2009). In addition to inner and outer sphere complexation, the surface-induced redox reactions could also play a role in the retarded migration of redox-sensitive elements (Geckeis, et al, 2013; Zavarin, et al, 2012). This is an important branch of the research of U, Pu and Np (Geckeis, et al, 2013). Another aspect to actinide-surface interactions is the biological material. According to Geckeis, et al. (2013) the actinide sorption on the microbial surfaces is not so different to inorganic surfaces e.g. clays. Similarly, fundamental mechanisms are taking place there also. The functional groups are more diverse compared to inorganic surfaces, which mainly consist of hydroxyl groups. The radionuclide intake by biomass should also be taken into account. For instance, the living cells can actively take up cations inside the cell via the metabolism which could affect the migration of certain radionuclides. (Geckeis, et al, 2013) 32

35 Europium chemistry and use as an analog for trivalent actinides Europium is the main radionuclide studied in the experimental part of this work. Therefore, its fundamental chemistry is an important aspect to be treated as a background information. Europium is an element with atomic number of 63 and it belongs to the group of lanthanides. In addition to Eu, there are 13 other lanthanides, 14 if Lanthanum is counted. The lanthanides derive their name from the latter, lanthanide meaning Lanthanum-like but the elements are quite usually called rare-earth metals. The term is somehow misleading since they are not very rare in nature. In contrast, they are often disturbing the radiological determination of actinides, which usually are appearing in very low concentrations in natural samples. This arises from the fact that the tri- and in some extent the tetravalent actinides and lanthanides are behaving chemically quite similarly, which makes the separation difficult. Especially the certain actinide/lanthanide pairs are very close to each other in chemical behaviour. One of these pairs is Eu/Am for example. Although it is sometimes disturbing the separations the analogy can also be utilized. Actinides are important regarding the long-term safety of final disposal of spent nuclear fuel. On the other hand, the actinides are sometimes difficult to research since they are not present in high concentrations or the handling requires heavy shielding to achieve sufficient protection. The analogy can help a lot. For example, Am behaviour is often researched by using Eu as an analog. Similarly, Gd can be used for Cm. Moreover, Eu is routinely used as an analog for all the trivalent actinides. (Lehto and Hou, 2011) The lanthanides are characterized by full K, L and M shells and 4s, 4p, 4d, 5s and 5p orbitals. They also have 4f electrons, the number growing as the atomic number increases. The lanthanides appear typically in oxidation state +III. Ce is an exception since it has only one 4f electron, which also participates in the formation of chemical bonds. That is why it also appears in oxidation state +IV. Eu has also a stable oxidation state +II. Remarkable with the lanthanides is that the 6p 2 and 5p 6 electron orbitals are shielding the 4f electrons. Therefore, the 4f electrons are not acting as outer-shell electrons meaning that they are not 33

36 forming any chemical bonds. The Ln 3+ ions are consequently formed by lose of 5s and 5p electrons. The shielding by 6p 2 and 5p 6 electron orbitals is causing a phenomenon known as lanthanide contraction. Usually the increasing number of electrons should also grow the size of the atom s electron cloud. With the lanthanides the shielding of 4f orbitals causes the electron cloud diminish as the atom number gets higher. This is caused by the growing nuclear charge, which increases the electrostatic attraction of electrons. Americium is an actinide and similarly to lanthanides, actinide means Actiniumlike. As stated before, the chemistry of actinides and lanthanides resembles each other. This originates to the shielded electron orbitals. With lanthanides the 6p 2 and 5p 6 electron orbitals are shielding the 4f orbitals. On the other hand, with actinides the outer 6d and 7s orbitals are shielding the 5f orbitals. This results in similar chemical behaviour. The chemistry of lanthanides is quite straightforward, essentially all of them are appearing in the oxidation state +III. In the case of actinides lighter than americium, the chemistry is much more complex. In addition to the outer electrons, also the shielded 5f are participating in the forming of chemical bonds. This leads to actinides appear in many oxidation states. For example, plutonium, the most complex actinide, can appear in natural conditions in oxiditation states +III, +IV, +V and +VI, even at the same time. Actinium and actinides heavier than plutonium are mostly appearing in the +III state. The interaction of Eu with bentonite colloids have been studied extensively (Schäfer, et al, 2004; Missana, et al, 2008; Bouby, et al, 2011; Schnurr, et al, 2015b). Missana et al. (2008) studied effect of bentonite colloids to the mobility of europium. They found strong sorption of Eu to the colloids in Grimsel Ground Water. In the same study, they found a strong effect of bentonite colloids to the mobility of Eu. Eu breakthrough from a column experiment with a granite fracture occurred at the same time with the conservative tracer and the bentonite colloids. No free Eu was detected. On the other hand, less than 7 % of injected Eu was recovered, which was far less than predicted in respect to the fraction of initially colloid-bound Eu. Missana et al. claimed this to be because of 34

37 reversible nature of europium sorption to bentonite colloids, which means that a fraction of initially colloid attached Eu was desorbed and retained inside the fracture. Similar conclusions have been made for Eu by (Bouby, et al, 2011) and for Am by (Kurosawa, et al, 2006). These results suggest that although Eu is strongly sorbed to bentonite, this interaction is reversible. This indicates that high sorption observed in simple laboratory setups cannot be interpolated straight to the more complex, natural systems. For example, the kinetics is an essential part of the colloid-associated migration of radionuclides (Missana, et al, 2008) 5. Theoretical background of experimental methods 5.1. Colloid Detection Laser-induced breakdown detection Theoretical background The laser-induced breakdown detection often shortened as LIBD or LIBS (Laserinduced breakdown spectroscopy) is a sensitive method for detecting particles in aqueous media but its applications are not limited to just that (Scherbaum, et al, 1996; Bundschuh, et al, 2000). For example polystyrene (Scherbaum, et al, 1996; Kitamori, et al, 1988), alumina (Scherbaum, et al, 1996) and clay mineral e.g. montmorillonite particles (Geckeis, et al, 2004; Delos, et al, 2008) can be effectively detected. The applications are not limited to these, as eigen-colloid formation of actinides has also been investigated with the method. (Bundschuh, et al, 2000; Bitea, et al, 2003) Also aerosols in gas phase can be analysed (Scherbaum, et al, 1996). The method relies on the generation of plasma by a pulsed laser which causes the matter to lose its dielectric property. An energetic pulsed laser beam is focused on a dielectric material, which detaches away electrons around the particles. This results in an electron avalanche and vaporization and ionization of 35

38 the material. There is certain value of the power density of the laser pulse, which has to be exceeded to achieve the breakdown. This is called the breakdown threshold value of the breakdown. The probability is the number of measured breakdowns detected divided by the total number of laser pulses applied for each measurement. The probability is not only dependent on the variable laser pulse density as stated before but also the particle concentration, size and material composition. Essentially, the probability is dependent on the number of the particle s weakest bound electrons, which are initially released. When the threshold is exceeded the atoms and molecules lose their electrons by photon absorption. The threshold of the solid materials is usually in the order of W/cm 2. Compared with water ( W/cm 2 ) and air ( W/cm 2 ) the threshold is much lower. (Kitamori, et al, 1988) That enables the controlled breakdown of the solid materials without the breakdown of the solvent if the energy of the laser pulse is adjusted correctly. Then the optical breakdown is produced only when a solid particle exists in the line of the beam. Remarkable is also that the small air bubbles found in the solution could pose a problem for the exact particle counting with laser scattering techniques like photon correlation spectroscopy (PCS), which will be treated later in this study. As the breakdown threshold is much higher for the air, the problem can be excluded with the LIBD. (Kitamori, et al, 1988) The electrons furthermore rip off the electrons from the nearby molecules, which results in a chain reaction, which culminate in the generation of plasma. The plasma is cooled down by the recombination of molecules and electrons and emission of excessive energy as light. The plasma is optically absorbed which follows up by a strong acoustic pulse. (Kitamori, et al, 1988) The schematic picture of the breakdown event is illustrated in Figure 5.1. (Bundschuh, et al, 2005a) Two remarkable effects result from the plasma: the emission of light and an acoustic wave. The emission of light can be used for the determination of the particle size and concentration as well as the elemental composition of the particles whereas the acoustic wave gives information only on the size and concentration of particles. (Scherbaum, et al, 1996) 36

39 Figure 5.1. The schematic pattern of the breakdown event. (Bundschuh, et al, 2005a) The signals are collected with computer-based image detection or a photoacoustic sensor, respectively. The light emission is recorded two- or threedimensionally by triggered Charge-Couple-Device Camera and the acoustic wave by a piezo electric detector. (Bundschuh, et al, 2005b) The plasma expansion produces a strong sound which can usually be heard by a human ear. (Walther, et al, 2002) Also gas bubbles are created as the result of the thermal expansion. When lighted up with an additional light source and analysing the shadow, information on the plasma itself can be obtained. (Walther, et al, 2002) The signals are amplified by voltage to current converters and finally the data is saved to a computer. The size of the particles is derived from the breakdown threshold and the mass concentration can be calculated from the slope of the energy curve. The breakdown probability depends on both particle size and concentration but in 37

40 principle, the breakdown threshold is affected just by the particle size. (Bundschuh, et al, 2005b) The effect of particle concentration can be seen in Figure 5.2 (a). With 110 nm particles, the breakdown probability is notable higher for three times higher concentration. The effect of the larger particle size can be seen in Figure 5.2 (b), where the breakdown probability is higher for 155 nm particles than for 19 nm. Another way to obtain the size information is the distribution of plasma events. The breakdown volume is larger for large particles compared to small particles. The recorded spatial distribution of the plasma events can be utilized as the extent of the distribution is directly dependent on size of particles. If particle size and breakdown probability are known, the colloid concentration can be derived. (Bundschuh, et al, 2005b) Finally, the signal is compared with the signal of the reference particles made e.g. of polystyrene and an average particle diameter and number density can be calculated. Figure 5.2. The breakdown probability curves of pure water and polystyrene particles with an excimer-dye laser (a) and Nd-YAG laser (b). (Bundschuh, et al, 2001b) The measurement of so-called s-curves allows measurement of size distribution of bimodal dispersions. The particle concentration is calculated by taking advantage of the breakdown probability with varied laser energy density. In order to obtain the particle size distribution, mathematical processing is needed. The product of s-curves is fit to the data by the MIGRAD algorithm of CERN-LIB MINUIT FORTRAN library. If not too much of large colloids are present, satisfactory good results can be obtained. (Walther, et al, 2006) The 38

41 measurement consists of sets of different laser pulse energies. The user determines the starting point and it depends on the materials investigated. The results contain the data of both the particle size and the concentration. The obtained s-curves can be analysed to derive the values for these variables. The example of s-curves of different particle concentrations and sizes are represented in Figure 5.3. (Walther, et al, 2002) Figure 5.3. The breakdown (BD) probability as a function of pulse energy. In the top picture, the effect of different particle concentrations can be seen. The threshold remains the same, but the slope is altered by the particle concentration. Below the threshold values can be seen decreasing with increasing particle size. (Walther, et al, 2002) As mentioned before, the laser pulse density should be chosen in a way that the matrix components would not undergo the breakdown. In the scope of the solids suspended in natural waters, the matrix components are the water molecules. If the threshold value of pure water were exceeded, the breakdown probability would in theory reach as an instant jump to 1, and all the breakdowns of solids would be lost in the saturated signal. Although this is not happening because of the impurities found even in ultrapure water and the alteration in the temporal profile of the laser pulse which causes the probability of breakdown events to rise as sigmoidal curve shape (see Figure 5.4) (Bundschuh, et al, 2001b). In other 39

42 words, the breakdown probability increases little by little until it reaches the saturation. By selecting the pulse energy below the threshold of the water, the colloids can be selectively detected. As said, the breakdown probability of the particles is dependent on both the concentration and the particle size. Figure 5.4 shows the dependency measured by the Bundschuh, et al. (2001b) Figure 5.4. The breakdown probabilities with the Nd-YAG laser as the function of particle weight concentration (a) and as function of particle number density (b) as taken from (Bundschuh, et al, 2001b). In Figure 5.4. (a), the breakdown probability is represented as particle weight concentration (Bundschuh, et al, 2001b). It should be noted that the breakdown probability is not higher for the same particle number (e.g. particle number density) with smaller particles. The contradiction is accounted for the particle weight concentration. For the same mass of larger and smaller particles, we naturally have much more small particles, which in proportion contributes to the higher breakdown probability. This can be seen in Figure 5.4. (b) (Bundschuh, et al, 2001b), where breakdown probability is represented as a function of the particle number density (ml -1 ). Noticeable is also that with increasing particle diameter, the breakdown probability is respectively rising. (Bundschuh, et al, 2001b) The advantage of the LIBD is its great efficiency with small particles and low concentrations. Although conventional light scattering methods like Photon Correlation Spectroscopy (PCS) can be efficiently used to detect particles with a 40

43 diameter of over 50 nm, the very low concentrations of the particles of this size are not detected. This is because of the difference in sensitivity of the methods. (Bundschuh, et al, 2001a) LIBD can easily determine concentrations in range of 1ppt ppms and the average size of nm. (Walther, et al, 2002) The limits of detection for PCS on the other hand, lies in concentration range between 100 ppt 100 ppm depending on the particle size. (Filella, et al, 1997) The exact limits of detection for LIBD are depending on the used system. As the different energy of excitation beam as well as different beam diameter and pulse duration can be utilized, a big difference in the sensitivity for example can be observed. Bundschuh, et al. (2001b) compared the method using two different laser sources: an excimer pumped dye laser and Nd-YAG laser system. The remarkable difference between the two systems was found related to the sensitivity of the method. For Nd-YAG laser, a minimum particle radius was 0.6 ± 0.3 nm, which is significantly lower than for an excimer pumped dye laser (9.8 ± 2.0 nm). Sample-preparation and measurement of the samples An essential part of all the particle size and concentration measurements is the clean working manners as the external particles could easily contaminate the sample vials and thus give unnecessary fluctuation to the results. LIBD measurements are not an exception to this: as the breakdown probability is rising with the size of the particles, the contamination of the samples e.g. by dust could lead to the loss of information on the analyte particles. (Walther, et al, 2006) Mostly ultrapure water with resistivity of 18.2 MΩcm and non-sorbing tubing and container materials should be used. (Walther, et al, 2002) The cells are made of quartz although several types of cells can be utilized. The type of cell depends on the volume of the sample. The cells should be cleaned with concentrated nitric acid following by rinsing with ultrapure water. (Walther, et al, 2002) 41

44 Ideally all the samples should be measured only once, since photon-induced chemical reactions might affect the results. This is achieved by using flowthrough cells. The sample in the focal volume is changed after a single laser shot. (Walther, et al, 2002) Different applications Many kinds of setups of LIBD have been used depending on the exact application. The LIBD has been used in laboratory experiments (Huber, et al, 2011; Huber, et al, 2015; Bitea, et al, 2003; Neck and Kim, 2001) but also in in-situ scale experiments where the online LIBD measurement system has been used. (Hauser, et al, 2002; Möri, et al, 2003a; Delos, et al, 2008) Applications in the nuclear sciences vary from the research of actinide speciation to the detection of pseudo-colloids, which could possibly act as carriers for radionuclides. The formation of thorium and other tetravalent actinide colloids has been studied by Neck et al (Neck and Kim, 2001; Neck, et al, 2001). A large number of peer reviewed articles about inorganic pseudo-colloids and their association with radionuclides, where the colloid detection has been done with LIBD, have been published. (Hauser, et al, 2002; Möri, et al, 2003a; Delos, et al, 2008; Huber, et al, 2011; Huber, et al, 2015) Apart from the nuclear sciences, the LIBD can be readily used for many applications. Routine quality control in variable industrial areas as food (Andersen, et al, 2016), steel (Li, et al, 2016) and semiconductor manufacturing (Kitamori, et al, 1988) is possible, as well as surveying the quality of wine (Bundschuh, et al, 2004), drinking water (Wagner, et al, 2003) or other kinds of natural waters (Walther, et al, 2006). However, this master thesis is focused on the inorganic colloids acting as carriers for radionuclides relevant to the final disposal of spent nuclear waste. 42

45 Photon Correlation Spectroscopy The photon correlation spectroscopy (PCS) is a powerful and relatively fast method for determining the particle size distributions and concentrations in aqueous media. It has advantages of being non-destructive and minimal alteration of the samples. (Holthoff, et al, 1997) It does not require any standards and it is suitable for size ranges from nano- to micrometer scale. (Filella, et al, 1997) The method is based on the analysis of the fluctuation of light scattering image produced by laser beam interaction with the particles in the fluid. The basic principle of size determination The suspended particles in fluid are not static in nature but they are constantly moving. This movement is called Brownian motion and essentially is generated because of solvent molecules hitting the particle surface. The Brownian motion causes particles to move all the time. The speed of movement although is dependent on the particle size. Since larger particles have larger mass they tend to move slower than the smaller particles if the chemical and physical composition is the same. (Malvern Instruments, 2007) When the particles are hit by the laser beam, the photon interacts with the electron field of the elements. This causes excitement of the electrons to higher energy level, which is soon relaxed by radiation. The radiation is emitted to all directions, in contrary to the strongly directional laser beam. These radiation photons are called scattered light, which is measured by a camera. Because the particles are moving the scattering image is changing over time. At t0 there is a certain scattering image. After a short time t1 the image has not changed much. If the two images are compared, a good correlation between them can be since the particles are almost at the same place as they were initially. If the time is extended, the image will change and thus the correlation decreases. After a long time the particles have moved far away from their initial position and there is virtually no correlation anymore. When this correlation data is collected with time-intervals, the correlation graph can be formed. The autocorrelation function is formed from the data and it is described in Equation 5.1 (Filella, et al, 1997) 43

46 lim 1 C( ) t I(t)I(t )dt Equation where I(t) = light intensity at time t = time interval = total time of measurement t1/2 = half-life of the specific nuclide. Two different autocorrelation functions are presented in Figure 5.5. The different pace of decrease of the correlation can be seen when comparing the two graphs. The smaller particles move faster resulting in a faster loss of correlation. This can be seen as a steep curve. In contrary, it takes longer time for bigger particles to move from their original position resulting in a milder fall of correlation. (Filella, et al, 1997) 44

47 Figure 5.5 Correlation functions for two different particle sizes. In the upper picture, the correlation remains longer. This indicates larger particles. Below, the decrease of the value of correlation coefficient is much sharper. The size distribution can be derived from this autocorrelation function. For dispersions with just one particle size and spherical particle form, the autocorrelation function decays exponentially with time. This can be described in the form of Equation 5.2. (Filella, et al, 1997) 45

48 g 1 (t) = exp(-dq 2 t) Equation 5.2 where D = diffusion coefficient q = wave vector The wave vector q can be described as Equation 5.3 (Filella, et al, 1997). q = 4pn l sinq 2 Equation 5.3 where n = refractive index l = wavelength of the laser beam q = scattering angle The diffusion coefficient, D, can be derived by fitting the experimental autocorrelation function to Equation 5.2. Further the hydrodynamic diameter, dh, can be determined from the diffusion coefficient by using the Stokes-Einstein equation (Equation 5.4.) (Filella, et al, 1997). 46

49 d h kbt 3pDh = Equation 5.4 where kb = Bolzmann constant T = temperature h = viscosity of the liquid Above only one particle size was assumed. The particle size is usually more likely a distribution than a single size in natural samples. Therefore, the autocorrelation function is a sum of the different particle sizes in the sample. The intensity of light is the initial parameter, which is measured in PCS. If the refractive index of the particles is known, the intensity distribution can be converted to particle number distribution by using Mie theory. (Filella, et al, 1997) It is important to notice that most of the mathematics involve the assumption of spherical particle size. In natural samples this is rarely the case, which results in distorted results. For example, the particles can be more like platelets or discs. The size distribution is in that way just suggestive although it may be possible to compare different particles with similar particle form. The proportionality of light counts to the concentration of particles can be utilized in the concentration determinations (Ledin, et al, 1993). More specifically, they have been utilized in the concentration determinations of bentonite colloids successfully (Missana, et al, 2008). For bentonite colloids, concentration range of ppm have shown linear correlation between the concentration and the count rate (see Figure 5.6). The authors state that multiple scattering encountered with higher concentrations prevents the quantitative measurements above this range. (Missana, et al, 2008) Similar experiments have been carried out by Niemiaho (2013). Within her study, she found at least 47

50 qualitative correlation between the count rate (PCS) and colloid concentration (ICP-MS) for bentonite colloids (see Figure 5.7). Figure 5.6. The light scattering counts of PCS measurement as the function of colloid concentration. The obtained calibration curve can be utilized in the calculation of colloid concentrations of suspensions. (Missana, et al, 2008) Figure 5.7. The light scattering counts of PCS measurement as the function of bentonite colloid concentration. (Niemiaho, 2013) Electrophoretic mobility Electrophoretic mobility measurements can be utilized to calculate zeta potential, which is an important parameter related to the solid surfaces in liquid media. Essentially it is related to the surface charge of the solids. If surface has negative zeta potential, it has negative net surface charge also. Vice versa applies for the positive zeta potentials. Moreover, it is linked to the particle repulsion or 48

51 attraction. With high zeta potentials, either positive or negative, the particles are repelling each other because of electrostatic forces. The attraction is at the maximum when zeta potential is close to zero since then the Van der Waals forces are dominating. It is important to keep in mind that it is not possible to directly measure the zeta potential, but it needs to be derived from some other measurable quantities. (Sonnefeld, et al, 2001) This can be done by measuring the electrophoretic movement, for example. In practice this is done by applying an electric voltage to the sample in a cell with two electrodes. The particles with a surface charge tend to move towards the opposite electrode by the velocity proportional to the magnitude of their charge. The velocity is measured by Laser Doppler Velocimetry. The zeta potential can be derived from the frequency or phase shift of the laser beam interacting with the moving particles. This conversion of shit to the zeta potential can be done either by Smoluchowski or Huckel theories. The principle of measurement is simple. The negatively charged particles move towards the positive electrode while the positive counter-ions move in an opposite way towards the negative electrode. The velocity of particle v has relationship with the electric field E, which can be expressed as Equation 5.5. (Luckham and Rossi, 1999) Ee re 0z v = Equation 5.5 h l where hl = viscosity of the liquid er = dielectric constant of solution e0 = dielectric constant of vacuum z = Zeta potential 49

52 We can also get the electrophoretic mobility e by Equation 5.6 (Luckham and Rossi, 1999). e r 0 l Equation 5.6 Further, the Zeta potential can be calculated if all the other parameters are known Radioactivity detection Within this work, the radionuclides emitting gamma and beta-energy were measured. A brief overview of the principles behind the methods are given below. Gamma radiation spectroscopy The gamma nuclides are mostly detected by scintillation crystal or semiconductor counter. The basic principle in both methods is quite similar: the energy of the photon is collected by a detector and its information is transformed into an electric signal. The scintillation crystals are often made of NaI but different materials, like CsI and Bi4Ge3O12 for example, can be utilized. (Klemola, 2002) The signal derived from the detector is amplified by preamplifier and is furthermore lead to the linear amplifier, which deforms the signal by making it shorter and stronger. After this modification, the signal is turned from analog signal to digital. The multichannel analyser separates the signals by their voltage and finally the computer calculates all the different signals. In the end, spectrum is obtained where usually the counts of energy range are shown as a histogram. In semiconductor detectors semiconductive materials like Ge or Si are used for the detector. Essentially no current is going through the detector if there is no excitation. When a photon hits the detector, a number of charge carriers is 50

53 created in the form of electrons and holes. This current is measured and the energy of photon is derived from the number of charge carriers. (Knoll, 2010) In the scintillation detection, photon causes excitation in the atoms of scintillator crystal. During the de-excitation, this energy is released as visible light and these scintillations are collected by photo multipliers. When the photon hits the crystal, the atoms are excited and visible light is emitted during the de-excitation. (Klemola, 2002) Liquid scintillation Beta radiation is routinely measured by liquid scintillation detection. Alfa and gamma radiation can be detected as well, but within this work the liquid scintillation was utilized just for beta radiation. The principle of the liquid scintillation detection is comparable to the use of scintillation crystal detectors. In the liquid scintillation detection, the scintillator is not solid material but liquid. Many different substances are used as scintillators. For example, anthracene and stilbene suit excellently to the measurement of radioactivity. The difference to the scintillation crystal is that the sample is solved straight to the scintillator. The scintillation process is although similar. Usually the scintillation cocktails contain two different scintillator substances: the primary and the secondary scintillators. The primary scintillator releases its excitation as a photon. The objective of the secondary scintillator is to the correct the wavelength to ensure an efficient multiplication process. (Ikäheimonen, et al, 2002) 51

54 6. Experimental Work In this master s thesis, the colloid-facilitated transport of radionuclides was studied by means of rock block fracture and supportive colloid stability and batch sorption experiments. The block-scale experimental set-up was used to simulate the repository conditions and to obtain knowledge of the interactions of radionuclides, montmorillonite colloids and the rock matrix. The primary objective was to research the effect of montmorillonite colloids to the migration of radionuclides. In addition, secondary objectives included the characterisation of the fracture, assessing its suitability for colloid-radionuclide migration experiments and developing the experimental setting for the work in future. Also, supportive batch sorption experiments were made with the rock and different montmorillonite materials to better understand the behaviour in larger context. The work was done at the Laboratory of Radiochemistry in the University of Helsinki, Finland and Karlsruhe Institute of Technology, Institute for Nuclear Waste Disposal (INE), Germany Materials Kuru Gray granite block The experimental set-up for a Kuru grey granite block having a natural water conducting fracture is illustrated in Figure 6.1. The fracture has been earlier applied to study the transport of iodide and sodium ion to evaluate the simplified radionuclide transport concept (Hölttä and Hakanen, 2002; Hölttä, et al, 2004; Hölttä, et al, 2008). The granite block originates from Kuru Quarry, Tampereen Kovakivi Oy, in Finland. The mineralogical composition of the fine-grained, non-foliated and equigranular rock consists of 38 % potassium feldspar, 21 % plagioclase and 8 % amphibole and mica minerals. The rest consists of e. g. quartz. The total porosity 52

55 and the surface area of the granitic material were studied by 14 C-PMMA method. The porosity was found out to be 0.2 %, while the density of the rock was 2660 kg/m 3. The block with dimensions of 90 cm x 90 cm x 70 cm has a horizontal fracture located across the block. For the hydraulic characterization, the fracture has been intersected by nine vertical drill holes having a diameter of 2 cm except the central drill hole diameter of 3 cm (Figure 6.2). The drill holes are equipped with sealing packers which prevent the water leakage out from the fracture. One of these sealing packers can be replaced with injection packer offering flexibility to change the flow channel. Fracture opening area around the block sides was covered with Plexiglas pools connected to each other to maintain the same hydraulic head inside the fracture. The additional water pools were installed below the fracture area and on the top of the block to ensure water saturation of the rock matrix and avoid capillary phenomena. For the collection of the tracer, the block was instrumented at the one outer vertical boundary where the horizontal fracture intersects the faces of the block. Figure 6.1. The Kuru grey granite block used with the work. The nine boreholes can be seen on the top of the block as well as the pools inserted on the sides of the block to allow proper saturation of the fracture. The lower pools and the upper pool maintain the water saturation of the rock matrix. 53

56 Hydrological properties of the fracture were characterised and the flow conditions and paths were determined previously by Hölttä et al. (Hölttä, et al, 2004). The hydraulic characterisation of the fracture was conducted by Hölttä et al. by forcing water to one borehole at time by applying constant head. Water inflow to a borehole was measured as a function of the hydraulic head applied in the borehole. Estimated transmissivities are illustrated in Fig. 6.2, which shows clearly the increase in transmissivity towards side 3. Based on the hydraulic and qualitative uranine dye tracer tests, the flow path from drill hole KR1 to side 3 was chosen for the tracer transport experiments. (Hölttä, et al, 2004) Figure 6.2. Locations of the boreholes (left) at the top of the block (Hölttä and Hakanen, 2002) and local transmissivities in a natural fracture (right) determined from the hydraulic tests (Hölttä et al. 2004) Water simulants Two groundwater simulants: low-salinity Allard water (I = M) and even less saline Grimsel Groundwater (I = M) were used in the experiments. Deionized MilliQ water (18 MW) was used for the preparation of groundwater simulants. Synthetic Allard water was used at first in the work. Allard reference water (Allard, et al, 1981) has been widely used as a standard low salinity granitic water simulant in the work done by Posiva Oy (Vuorinen and Snellman, 1998; 54

57 Hellä, et al, 2014) for example. The composition of the Allard water is shown in Table 6-1. Table 6-1. The composition of Allard Simulant. The final ph of the solution was adjusted to 7 by adding 0.1 M HCl. Component c c [mg/l] [mmol/l] MgSO 4 7 H 2O MgCl 2 6 H 2O KCl CaCl 2 2 H 2O Na 2SiO 3 9 H 2O NaHCO The four first components were added from the stock solutions prepared forehand by weighing substance to a volumetric flask and then adding MilliQ water to achieve the intended volume. Then Na2SiO3 9 H2O was added as a freshly prepared solution, which ph was adjusted to 8 by adding 0.1 M HCl. The ph adjustment of prepared Na2SiO3 solution was found to be important since without it white precipitation occurred during the addition to the final water solution. Then weighed mass of NaHCO3 was added straight to the volumetric flask where the final solution was to be prepared and ph was adjusted to 9.3 by use of NaOH. The chemical composition of Grimsel Groundwater was gathered from various sources in literature (Aksoyoglu, et al, 1991; J. Eikenberg, et al, 1991; Frick, et al, 1992; Schäfer, et al, 2013) and the recipe was formed based on the information provided in these papers. Also the publications, where synthetic Grimsel Groundwater Simulant were used, were utilized when forming the recipe. (Muuri, et al, 2016)The composition of the used Grimsel Groundwater simulant is shown in Table 6-2. The preparation of water simulant, as well as all the stages of the experiments, was done in the ambient laboratory air. For this reason, it is likely that the CO2 intrusion has taken place, which affects the chemical behaviour of the system. On the other hand, the ph was measured in many 55

58 stages and no significant change was observed in the water simulants. This indicated that the system was stable at least before the contact with the rock. Table 6-2. The recipe of the Grimsel Ground Water Simulant used in the work. Component c c [mg/l] [mmol/l] Ca(NO 3) 2 4 H 2O HF (40 %) H 2SO NaHCO NaCl Na 2SiO 3 9H 2O Montmorillonite colloids Two kinds of montmorillonite colloids were used in the experiments. First montmorillonite was the synthetic Ni-incorporated montmorillonite (Figure 6.3) which was obtained from KIT-INE. Quite often natural, purified montmorillonites are used on the experiments related to radionuclide migration. Two elemental indicators for the chemical analysis with ICP-MS for example are Si and Al, the major components of the montmorillonite. However, the detection of Si suffers from a high background because of the materials of the ICP-MS device, which is the main reason for the use of Al as a main indicator. On the other hand, in experiments with low colloid concentrations the chemical analysis might be difficult because of low elemental concentrations compared to the background concentrations. That poses a need for a synthetic material, where another kind of elemental indicators are incorporated into the structural matrix. These elemental indicators are for example Zn and Ni, from which the latter was used in this study. The montmorillonite mineral powder had been synthesised by the staff of KIT-INE as described by Reinholdt, et al. (2001, 2013). Briefly, Nimontmorillonite had been synthesized in acidic and fluoride medium under 56

59 hydrothermal conditions and it was characterized with X-ray diffraction. Its formula per half unit cell without tetrahedral substitution is Na0.3[Al1.64Zn/Ni ] [Si4.00]O10(OH1.95F0.05) 4.0 H2O, where stands for the vacancy. (Huber, et al, 2015) The structure of Ni-labelled montmorillonite is presented in (Figure 6.4). Figure 6.3. Centrifgutaed Ni-labelled montmorillonite suspension. A characteristic colour is clearly visible. Figure 6.4. The modelled structure of Ni-montmorillonite. Ni atoms are the blue coloured diamonds while Al are shown in pink colour. The interlayer water can be seen between the two montmorillonite layers. (Puhakka, 2016) Another type of montmorillonite used within the work was manufactured by Nanocor Company. The product name Nanocor PGN is purified montmorillonite 57

60 (>98 %). The montmorillonite was provided by B+Tech, who had also done the characterization. (Kiviranta, et al, 2016) Three different colloid suspensions were prepared for this study. Two suspensions were prepared for Ni-montmorillonite: one with Allard water and one with purified MilliQ-water. Only the MilliQ was used with the suspension prepared from Nanocor montmorillonite. The colloid suspensions were prepared by adding 1-2 g of montmorillonite powder per 100 ml of water. The preparation procedure consisted of cycles of water changes. The separation was done by centrifugation at 4000 RPM (1.935 x G) and the supernatant was removed by decantation. After this the solid was re-suspended by shaking it vigorously and then putting the vial to the automated shaker for a longer period. The time between the separation and re-suspending differed between the stages of preparation. The first equilibration time was approximately ten days, while the three others were around three days. The equilibrating schedule was quite similar to the work done by Huber, et al. (2015). The samples were taken between every stage and they were analysed by measuring the conductivity as well as the particle size distribution with Photon Correlation Spectroscopy (Malvern Zetasizer Nano ZS). The final stock solution was also characterised by ICP-OES to determine the final concentration. Al, Si and Ni were used as indicators of the concentration of montmorillonite. The montmorillonite concentrations could be calculated from the element concentrations when the mass fractions of the elements were known. These can be calculated from the composition of the unit cell. The Ni-montmorillonite colloid suspension was prepared by adding 0.5 g Nimontmorillonite to 50 ml of water. The first equilibration lasted for 11 days, second 3 days, third 4 days and fourth 3 days. The Nanocor colloid suspension was prepared similarly to Ni-montmorillonite suspension. 2.0 g of Nanocor PGN Montmorillonite powder was suspended to 120 ml of MilliQ-water. The centrifuge, the speed and time parameters were the same (4000 RPM, 40 min). Equilibrating times were five days for the first 58

61 equilibration, three days for the second and four days for the third. Similarly, samples were taken in each of the stages and the conductivity and the size distributions were measured. The final solution was also measured by MP-OES to determine the final colloid concentration. A picture of prepared Nanocor montmorillonite solution is shown in Figure 6.5. Figure 6.5. Freshly prepared Nanocor PGN montmorillonite suspension after the fourth centrifugation step. After this the supernatant was moved to another vial and it was used for the experiments. Tracers Conservative dye and radioactive tracers were used within the work to determine the water flow properties in a fracture. The dye tracers were Amino-G (AGA), 7-Amino-1,3-naphthalenedisulfonic acid monopotassium salt monohydrate, purchased from Sigma-Aldrich and sodium fluorescein (uranine) purchased from Merck. The dye tracer solutions were made by weighing certain mass of the 59

62 analytical grade colouring agent to a vial and then dissolving in MilliQ water and adjusting to the desired concentration. The prepared tracer cocktails are identified in Table 6-3. Table 6-3. Non-radioactive tracer stock solutions prepared for the work. Solution ID m(dye) m(total) c(dye) c(dye) c(dye) [g] [g] [g/l] [mmol/l] [ppm] Fluorescein Amino-G The non-sorbing radioactive tracers were 3 H and 36 Cl used as a tracer cocktail or in the experiments as just 3 H solution, which were prepared by adding desired volume of the stock solutions. The initial stock solutions had activity concentrations of 169 kbq/ml ( ) for 3 H and 26 kbq/ml ( ) for 36 Cl. No specific information on the carrier content was available for the stock solutions. There was probably carrier added to the 36 Cl solution but this could not be confirmed. The tracer solutions were made by pipetting a certain volume of the stock solution corresponding to the amount of radioactivity desired for the final solution. Then the solution was diluted with MilliQ water to achieve the intended activity concentration and volume. The first non-sorbing radioactive tracer solution contained radionuclides 3 H and 36 Cl. The second contained only 3 H. The activity concentrations of the tracer solutions are shown in Table 6-4. Table 6-4. Conservative radioactive tracer solutions The radioactivity has been decay-corrected for the date of the preparation. Solution ID As ( 3 H) [kbq/ml] As ( 36 Cl) [kbq/ml]

63 The sorbing 85 Sr and 152 Eu tracers were purchased from Isotope Products Laboratories, USA. The activity concentration of initial 85 Sr solution was 3.7 MBq/g ( ). The information about the concentration of carrier was not given by the manufacturer. Since the manufacturer had not determined the activity concentration of the original solution, the prepared stock solution was measured with a semiconductor detector (Genie 2000) in a known and calibrated geometry to ensure the radioactivity concentration, which was determined to be 880±40 kbq/g. The 152 Eu tracer used for the batch experiments at KIT-INE was prepared from a highly active stock solution, which was a carrier-free 10 MBq / ml (01/2014) solution from which a dilution of 50 kbq / ml had been prepared previously. This solution was used for the batch experiments conducted at KIT-INE Methods UV-VIS spectrometry The concentration of two colouring dyes were determined using Perkin Elmer Lambda 40 UV-VIS spectrometry. For the fluorescein, the wavelength of 491 nm was used. For the Amino-G, the wave length of 310 nm was used at the beginning according to the manufacturer s reference. The absorbance measurements with different wavelengths gave better sensitivity for 250 nm which was chosen for the rest of the measurements. Samples were measured in Quartz-cuvettes initially using volume of 2 ml which was increased to 3 ml due to the more optimal measurement geometry. The device was calibrated with the background electrolyte. A standard series was measured each time before the set of samples to make sure that the device was stable. 61

64 Photon Correlation Spectroscopy Photon correlation spectroscopy (PCS) was used to characterize the colloid suspensions used in the work. The device used in UH was Malvern Zetasizer Nano ZS (Figure 6.6) and in KIT-INE ZetaPlus system manufactured by Brookhaven Instruments Corp. Figure 6.6. Malvern Zetasizer Nano ZS. The device was used to determine the particle size distribution in the solutions. Malvern Zetasizer Nano ZS is able to measure zeta potential as well as the particle size distribution in the approximate range of nm. Particle size was measured using a glass cuvette (8G) filled with 1 ml of solution. Zeta potential was measured using the dip cell, which was immersed in the glass cuvette used in the size measurements. (Figure 6.7). 62

65 Figure 6.7. The glass cell used for the particle size distribution measurements. The inserted dip cell electrode for Zeta potential measurements can be seen on the top of the cell. Laser induced breakdown detection Laser induced breakdown detection (LIBD) is an effective method to determine colloid size and concentration. Compared with PCS, LIBD is many magnitudes more sensitive for colloid characterisation, especially in the case of small colloids (< 100 nm) (Bundschuh, et al, 2001a). Therefore, it is suitable for the low particle concentrations relevant to natural conditions. The LIBD measurements were done at KIT-INE using the optical LIBD device (see Figure 6.8), which had been previously calibrated with polystyrene particles with known particle sizes. The schematic figure of the used system is shown in Figure 6.9 The laser was Continuum Minilite Nd:YAG laser, which was operated on 203 mv voltage. The voltage adjustment was done by hand and the energy control was monitored via Tektronix TS220 oscilloscope. There was also a control unit and water based cooling unit attached to the laser. The breakdown events were observed by using Progressive Scan CV-M10 CCD camera with Leica 63

66 Apozoom 1:6 objective. Finally, there was a computer doing the actual data recording. Before each use the laser was put on and it was left to warm up for at least half an hour. Even after that it was necessary to monitor the laser energy as the variation affects enormously the breakdown probability and so on the particle concentration results. The spatial distribution of the breakdown events is collected and from these data the particle size distribution and concentration are calculated. Figure 6.8. LIBD device. (Insitute for Nuclear Waste Disposal, 2015) Another critical requirement is the purity of the cuvettes. This was achieved by rinsing the cuvettes with a large volume of MilliQ water. The purity was checked by measuring cells filled with MilliQ qualitatively with LIBD. The alignment was to get breakdown probability of pure MilliQ below 10 %, preferably around 5 %. Usually breakdown probability of % was tried to reach. The breakdown probability with samples containing no colloids was still lower because no particles were detected. Usually sample dilution varied between 1:1 and 1:

67 Figure 6.9. A sketch of the optical LIBD system used in Karlsruhe. Radioactivity Measurements The gamma activities were measured using two different systems. The Canberra Ge-detector with Genie 2000 software was used to measure activity of the original 152 Eu tracer solutions. Also, some samples from the block experiments were measured using Ge-detector since more sensitive method was needed at times. The most of activity measurements for 152 Eu and 85 Sr were done with 1480 Wizard 3, which have been manufactured by PerkinElmer Wallac. The similar device was used in Helsinki and Karlsruhe. They have a NaI crystal with dimensions of 80 mm x 75 mm. The devices were connected to computer where the data was processed. A blank sample was used to subtract the background counts from the actual counts of the tracers. The activity results were compared to a known standard or the initially injected activity. The radionuclides without gamma emission, 3 H and 36 Cl, were measured with liquid scintillation. The main device used for the measurements was Hidex 300 SL. A quench curve was made for 3 H with low beta energy to correct the counting efficiency with the quench parameter TDCR. No similar measurements were 65

68 done for 36 Cl with higher beta energy, which is not be quenched in such a remarkable way. Altogether, the exact activities of 36 Cl and 3 H were not a crucial part of this study since the two tracers were just used to check whether the Amino-G is acting similarly to these two. MP-OES Both MP-OES and ICP-OES were used during the work. The MP-OES was used in UH by the author, while the ICP-OES was used at KIT-INE and the measurements were done by person responsible of the device. The device in UH was Agilent 4200 Microwave Plasma-Atomic Emission Spectrometer. The procedure for the sample preparation was rather simple. At KIT-INE, the measurements were done with HF background, where the sample was dissolved. The procedure in UHwas similar but with use of different acid: the samples were diluted with HNO3 to achieve a 5 vol-% HNO3 solution. Al, Si and Ni were used as probes for montmorillonite colloids since these can be found in its chemical structure. The same probes were used in Helsinki and Karlsruhe, but in latter also Na was measured. In Helsinki, the system was calibrated by having standard series of six samples. For Al the range of 0-10 ppm (0, 1, 5, 10, 15, 25 ppm) was used. Correspondingly for Si the range was 0-40 ppm (0, 10, 25, 40, 75, 100 ppm) and for Ni 0-8 ppm (0, 1, 5, 10, 15, 25) range was used. The standard series samples were prepared by adding commercial standards to HNO3 and finally solutions were diluted to come up with 5 vol-% HNO3 solutions. The measurement parameters have been presented in Table 6-5. Table 6-5. The most important measurement parameters for MP-OES in Helsinki. Also other wavelengths were measured but these were chosen because they suffered least of disturbance. Parameter Value Probe Wavelength [nm] Stabilization time 15 s Si ; Sample uptake 25 s Al ; Rinse time 30 s Ni ;

69 6.3. Experimental settings Block scale experiments The block scale experiments were done by injecting water simulant to the block fracture and collecting it from the side pool. A spike of tracer was injected and the appropriate methods were used for detecting the tracer. The injection system consisted of a plastic water drip, a peristaltic pump and an injection packer (Figure 6.10). Initially a 50 l injection loop was used for the tracer injection but soon it was abandoned because of low reproducibility. The next generation injection system consisted of a three-way valve, which was between the drip and the pump. The additional hose could be put inside a vial containing the tracer. This way also bigger volumes could be injected and thus lower concentrations used. Also, the reproducibility was better. The injected mass was obtained when the vial containing the tracer was weighed before and after the injection. Some water was injected via the same hose used for the tracer injection after the injection. This way the dead volume was rinsed with a volume many times higher than the volume of the hose and the full injection assured. The drill hole KR1 was used for the injection of water inside the fracture. The nine boreholes are shown in Figure One can see the injecting hose going inside the hole KR1 on the bottom right of the picture. 67

70 Figure The injection system consisted of a water drip, peristaltic pump and a packer located inside the drill hole. In the first tests also an injection loop was used but soon this was replaced by a different system because of reproducibility problems. Figure The block from above. The nine boreholes can be seen from which the hole KR1 (on the bottom right) was used for the injection. The sampling was done on the other side of the block. The collection consisted of two collection areas, which were separated from the rest of the pool by metal rods inside silicone hoses, and two automated sample collectors. A picture of 68

71 sample collection system is shown in Figure The samples were collected with another peristaltic pump. Different analysis methods were used to analyse the samples depending on the analyte: gamma and liquid scintillation for radioactive tracers and UV-VIS for the colouring dyes. Also, qualitative detection was utilized by using an UV-lamp (Figure 6.13) as the fluorescence was then visible without any instrument. This way the injection process could also been followed visually. Figure The sample collection system. The collection areas can be seen on the right part of the facing block wall. Two automated sample collectors are shown in the front and the peristaltic pump can be seen behind a plastic canister. The canister was used to pump water to the collection pool to compensate the volume loss from sample collection and evaporation. This was done to keep water level constant and to ensure that the fracture remained below the water surface all the time. Figure The block samples in UV-light. Even quite low concentrations were able to see with the light, although the detection limit was not determined this way furthermore. 69

72 Batch experiments Batch experiments were conducted to gain supportive data on the colloid behaviour with similar granite material as the one used in block experiments. As the block scale experiments are and appeared practically to be complicated and difficult to accomplish, more simplified experiments were needed. The initial objective was to study the ternary system similar to block experiments with ground water (Grimsel Ground Water Simulant), granite (same Kuru Gray material as the block) montmorillonite colloids (both synthetic Ni-labelled and Nanocor PGN) and a sorbing tracer ( 152 Eu). The experiments concentrated on the reversibility of the Eu-sorption as well as its kinetics. In general, the idea was to investigate where does the Eu end up and in which ratios: is it in colloids, in solution or in the rock material. The colloid fraction was removed with centrifugation. At UH the centrifugation was done using RPM and at KIT- INE using RPM. Quite a difference can be seen between the two centrifugation speeds, which was because of the ultracentrifuge was under maintenance at KIT-INE. Therefore, normal centrifuge had to be utilized. The succeeding of the separation was however controlled either with PCS or LIBD. The dissolved or free Eu fraction was measured from the supernatant. The Eu fraction in rock material was calculated when the preceding was subtracted from the initially added activity. The detailed sample preparation was done as follows. In the phase one 1 g of MilliQ-washed crushed Kuru Gray granite was put to a 20 ml scintillation vial with 9 ml of Grimsel Ground Water Simulant. The material and water was equilibrated for 9 days, including two water changes, when the solution was decanted and replaced with a fresh simulant. Samples for analysis were taken during both water changes. After 9 days, the water was changed again. This time the replacement was done with the tracer solution containing 152 Eu and montmorillonite colloids, which had been equilibrated for one day. The scheme of the samples preparation for Ni-montmorillonite is shown in Figure Two different sample series with different particle sizes ( mm and 1-2 mm) were done with two parallel samples for each sampling point. 70

73 Figure The scheme for preparation of samples for the batch experiments. First granite granules were rinsed. After that they were equilibrated with water. After the equilibrium process the water was changed to GGW solution with montmorillonite colloids and 152 Eu tracer. The sampling was done with intervals. The sampling procedure for individual samples is shown in Figure The similar procedure was used with the batch experiments conducted in Helsinki. In these experiments, Ni-montmorillonite was changed to Nanocor PGN and just the size fraction mm was used. The colloid detection was done with PCS instead of LIBD. 71

74 Figure The sampling procedure for samples of batch experiments. First samples were taken from liquid phase for LIBD, gamma counting and ICP-MS. Another 4 ml sample of liquid was centrifuged and samples were taken for the same measurements. Then ph was measured from the liquid on the top of the granite. Finally, all the liquid was removed and replaced with fresh water. This water change was done to see if the sorption of Eu or attachment of colloids was reversible. This would be seen as the increase in radioactivity or particle concentration. 72

75 Eu radioactivity was measured directly from the solution phase with and without colloids. The fraction remaining in the granite was calculated with Equation 6.1 based on these values. c granite = A 0 A solution A 0 Equation 6.1 where A0 = radioactivity of Eu in initial solution Asolution = radioactivity of Eu in non-centrifuged sample The proportional colloidal fraction was calculated using Equation 6.2. c colloids = A solution A centrifuged A 0 Equation 6.2 where Asolution = radioactivity of Eu in non-centrifuged sample A0 = radioactivity of Eu in initial solution Acentrifuged = radioactivity of Eu in centrifuged sample The dissolved fraction was calculated using Equation 6.3. c dissolved = A centrifuged A 0 Equation 6.3 where Acentrifuged = radioactivity of Eu in centrifuged sample A0 = radioactivity of Eu in initial solution 73

76 7. Results and discussion 7.1. Colloid characteristics Ni-montmorillonite suspensions for batch and block experiments were prepared in Allard and MilliQ waters by suspending mineral powder in water and changing the solution for four times. In each stage the separation of solid and liquid material was done by centrifugation and after that the solution was changed to fresh water. The preparations were done under ambient laboratory air. The characterisation of prepared suspensions in each stage consisted of conductivity, particle size distribution and Zeta-potential measurements. The aim of the characterisation was to ensure that the prepared colloids were stable. The results of these measurements are shown on the next section for each separation stage. The corresponding values for Zn-montmorillonite, which had been previously prepared and measured at KIT-INE, are also given for comparison on some graphs. Zn-montmorillonite is similar material to the Nimontmorillonite but Zn is added to the montmorillonite structure instead of Ni. The measured conductivities of Ni-montmorillonite in MilliQ and Allard are shown in Figure 7.1. The conductivities were much higher in Allard than in Grimsel groundwater or MilliQ which is explained by the four times higher ionic strength of Allard compared to GGW. Even though MilliQ is ion-free, the conductivity was almost 100 μs/cm after the first centrifugation cycle because of the dissolved salts from the montmorillonite. The reason for decreasing conductivity after the next centrifugation cycles is that most of the salts were dissolved during the first suspending stage. 74

77 Conductivity [us/cm] Ni-mont in Allard Zn-mont in GGW Ni-mont in MilliQ Centrifugation Cycle Figure 7.1. The conductivities of Ni-montmorillonite suspensions in MilliQ and Allard water and Znmontmorillonite in Grimsel groundwater after sequential centrifugation cycles. Mean particle size is an important parameter in the research of colloidal particles as the stability of colloids is directly related to the particle size. Particle size distributions measured with PCS are shown in Figure 7.2. The mean particle sizes for the two lowest ionic strength solutions, Ni-montmorillonite in MilliQ and Zn-montmorillonite in GGW were similar being between 200 and 300 nm in every centrifugation stage. The particle sizes were in the colloidal region, i. e. below 1000 nm, where colloids are assumed to be stable and mobile. The particles in Allard were much larger being around 500 nm during the two first stages and ending up in values over 700 nm after the second centrifugation cycle. The growing particle size suggests that the Ni-montmorillonite colloids are not stable in Allard. It should be stated that the particle geometry is approximated to be spherical in PCS measurements, which is not appropriate for smectites, e.g. montmorillonite. This should be taken into account when discussing about the exact colloid sizes. 75

78 Particle diameter [nm] 1000,0 900,0 800,0 700,0 600,0 500,0 400,0 300,0 200,0 100,0 0, Centrifugation Cycle Ni-mont in Allard Zn-mont in GGW Ni-mont in MilliQ Figure 7.2. The particle diameters of Ni-montmorillonite colloids in Allard (blue diamonds) and in MilliQ (purple triangles) after sequential centrifugation cycles. Particle sizes of Zn-montmorillonite in Grimsel ground water (red squares) are given for the comparison. The standard deviations of different measurements are shown as error bars. For most of the data points standard deviations were so small that they hide behind the symbols in the graph. Zeta potential is another parameter related to the colloid stability which is related to the surface charge of the colloids. Values close to zero indicate instability of colloids. Zeta potentials of Ni-montmorillonite colloids in Allard and MilliQ waters are shown in Figure 7.3. Zeta potentials are rising from -35 to -20 mv in Allard but falling from -35 to -45 mv in MilliQ. Unfortunately, no data on the behaviour of the Zn-montmorillonite was available for the comparison. 0 Zeta potential [mv] Ni-mont in Allard Ni-mont in MilliQ Centrifugation Cycle Figure 7.3. The zeta potentials of the particles in Allard (blue diamonds) and in MilliQ (purple triangles) after sequential centrifugation cycles. The standard deviations of the repeated measurements are shown as error bars. 76

79 Derived count rate is the PCS measurement parameter which is related to the particle concentration. It is an attenuation normalized number of light scattering events counted per time and higher count rate indicates higher particle concentration. It should be noted that the count rate is also affected by the size of the particles so one should be careful when comparing count rates of particles of different size distributions. The derived count rates measured for each stages of prepared Ni-montmorillonite suspensions are shown in Figure 7.4. On the contrary to conductivity and Zeta potential measurements, no such a clear difference could have been found in the behaviour of Ni-montmorillonite colloid concentration between Allard and GGW. The evolution of count rates was similar in both solutions and the particle concentrations decreased after the first centrifugation cycle. Derived Count Rate [kcps] Centrifugation Cycle Ni-mont in Allard Ni-mont in MilliQ Figure 7.4. Derived count rates of Ni-montmorillonite suspensions in Allard (blue diamonds) and in MilliQ (purple triangles) after sequential centrifugation cycles. Derived Count Rate is a parameter proportional to the concentration of the particles. The stability of Ni-montmorillonite suspensions was followed in two different GGW solutions by measuring the samples using PCS. First solution was freshly prepared GGW, second was GGW, which had gone through the granite block fracture (Breakthrough GGW) and two parallel samples were prepared for both solutions. The objective was to ensure the suitability of colloids for the block scale experiments. GGW, which had gone through the block, was used to ensure that the colloids are behaving similarly in conditions inside the block fracture in comparison to fresh GGW. The results of ph-measurements are shown in Figure 7.5, conductivity measurements in Figure 7.6, the average particle sizes in Figure 77

80 7.7 and the derived count rates in Figure 7.8. The stability of colloids in one of two GGW samples differed from the others. The blue circle (GGW1) sample shows growing particle size (see Figure 7.7) and remarkable increase in conductivity (see Figure 7.6). This could be because of increasing ionic strength of the solutions, which could originate for example in spillage of ph-electrode. That could considerably increase K +, as potassium chloride electrode was used, content of the solution and thus increase the conductivity of the solution. No remarkable differences could be found between the other samples. Results show that colloids remained stable for at least the duration of the experiment and they are stable in GGW prior and after the contact with the rock material to be used on the block scale experiments. ph 9,0 8,5 8,0 7,5 7,0 6, Time from t0 (h) Conductivity (ms/cm) 2,50 2,00 1,50 1,00 0,50 0, Time from t0 (h) Figure 7.5. The ph values of montmorillonite colloids in fresh GGW (circles) and in GGW extracted from the granite fracture (diamonds). The different colours of the same symbol refer to parallel samples. Figure 7.6. The conductivities of Nimontmorillonite suspensions in fresh GGW (circles) and in GGW extracted from the granite fracture (diamonds). Z-ave (nm) Time from t0 (h) Derived Count Rate (kcps) Time from t0 (h) Figure 7.7. The average particles sizes of Nimontmorillonite colloids in in fresh GGW and in GGW extracted from the granite fracture. Figure 7.8. Derived count rates of Nimontmorillonite suspensions in fresh GGW and in GGW extracted from the granite fracture. Derived Count Rate is a parameter proportional to the concentration of the particles. 78

81 The derived count rates were used to calculate montmorillonite colloid concentrations in the prepared suspensions. The standard series (see Figure 5.7) was used to calculate the concentrations. The concentrations and other parameters measured for the colloid suspensions are shown in Table 7-1. Z-ave stands for the intensity based average size of colloids, number mean is the number based mean size of the colloids, derived count rate is the light scatter counts per time and PdI is Polydispersity index, which is related to the polydispersity of the sample. Values close to zero indicate that all particles in the sample are of the same size. On the contrary, values close to 1 indicate that particles in the sample have a very broad size distribution. Table 7-1. The essential parameters determined for the prepared colloid suspension by PCS. NiMW stands for Ni-montmorillonite material, followed by the type of water used. Nanocor suspension was prepared only in MilliQ. Stock solution Number Derived Z-Ave [nm] Mean [nm] Count Rate [kcps] PdI Conc. [mg/l] NiMW_Allard NiMW_Allard NiMW_MilliQ NiMW_MilliQ NiMW_MilliQ Nanocor_MilliQ The elemental compositions of montmorillonite suspensions were also analysed with ICP-OES (KIT-INE) and MP-OES (UH). The results of elemental analysis are shown in Table 7-2. The colloid concentrations were calculated from the elemental concentrations by taking into account the relative abundancies of each element in the structure of montmorillonite. It seemed that the calculated concentration varied noticeably comparing with the values calculated from the results of PCS (see Table 7-1). This indicates that the standard series made with different kind of bentonite may not be suitable for the concentrations calculations with another kind of material, although they are both montmorillonite. It seems that the results from PCS and MP-OES are rather in 79

82 line for Nanocor (3700 and 4100 mg/l, respectively) but not for the Nimontmorillonite (330 and 30 mg/l, respectively). The results from PCS and ICP- OES ( and 150 mg/l) seem to differ but not as much as the results between PCS and MP-OES. The reasons for the differences remain uncertain, but the answer could be different size distributions between Nanocor and Nimontmorillonite colloids. Further experiments are required to gain better understanding of the differences between the different montmorillonite colloids and their analysis. Table 7-2. The elemental compositions of different colloid suspensions and the calculated colloid concentrations. The uncertainties are calculated from the standard deviations of the independent measurements for different elements. The uncertainty for the final colloid concentration is calculated from the standard deviation of the colloid concentrations calculated from the concentrations of different elements. All the units are mg/l. Stock solution NiMW_MilliQ1 (KIT-INE ICP-OES) NiMW_MilliQ3 (UH MP-OES) Nanocor (UH MP-OES) Si Colloid Conc Al Colloid Conc Ni Colloid Conc Na Colloid Conc Average colloid conc Eu distribution between the granite, colloids and ground water Batch sorption experiments on the behaviour of Eu in a system including granitic rock, montmorillonite colloids and groundwater were conducted to estimate the Eu distribution between the different phases. The series of samples were prepared and two parallel samples were analyzed at each sampling point. The distribution of Eu was followed by measuring the radioactivity by gamma spectroscopy. The samples were also measured with LIBD or PCS and ICP-OES to follow the colloid stability. ph was also measured at every sampling point. 80

83 Experiments were done with two montmorillonite colloids: the first being synthetic Ni-montmorillonite and the second Nanocor PGN montmorillonite. Grimsel ground water simulant was used as the solution. Two grain sizes of granite, and 1-2 mm, were used for the experiments done with Nimontmorillonite. Only grain size mm was used on the experiments with Nanocor PGN. The proportional 152 Eu distribution in solution, on Ni-montmorillonite and granite fracture material are shown in Figure 7.9 and in Figure The two graphs refer to two different experiments conducted with different mesh size of granite, the first having mesh size of mm (Figure 7.9) and the second 1-2 mm (Figure 7.10). Quite similar behaviour of Eu was observed in the experiments for the both grain sizes of the granite. The percentage of Eu in solution (blue diamonds) was % (grain size mm) and % (grain size 1-2 mm) for all the samples. This remained rather constant during the experiments. The colloid-bound Eu fractions were initially lower in the samples with granite material (green triangles) compared to the solution containing only montmorillonite colloids and Eu (purple line). This indicates that sorption competition is occurring immediately after the contact with granite. Later, the colloid-bound Eu fractions were getting even lower and at the same, the granitebound fraction increased (red squares). Although as mentioned, no increase in free Eu concentration in solution can be seen (blue diamonds). Consequently, it seems that the Eu desorbed from colloids and attached to the solid granite. 81

84 152 Eu c/c0 1 0,9 0,8 0,7 0,6 0,5 0,4 Eu attached on colloids before contact with granite Colloids Granite 0,3 0,2 0,1 Dissolved t [h] Figure 7.9. The proportional 152 Eu distribution in GGW solution (blue diamonds), on Nimontmorillonite colloids (green triangles) and mm granite (red squares) as the function of time. The purple line shows the percentage of Eu attached on colloids before the contact with granite. 152 Eu c/c0 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 Eu attached on colloids before contact with granite Colloids Granite 0,2 0,1 Dissolved t [h] Figure The proportional 152 Eu distribution in GGW solution (blue diamonds), on Nimontmorillonite colloids (green triangles) and 1 2 mm granite (red squares) as the function of time. The purple line shows the percentage of Eu attached on colloids before the contact with granite. 82

85 The reason for the observed decrease in the colloid-bound Eu fraction could be the desorption of Eu from colloids following sorption to the granite or polyethylene vial or the instable colloids. The extent of sorption to the vial is difficult to confirm since no activity of the solid material was directly measured. The ph data (see Figure 7.11) shows that ph did not remain stable. The initial ph of the water used was 9.3 (GGW) but it lowered to 8 after the contact with granite. After the first sample, it increased to 8.5. One of the two samples for mm (blue diamonds) at 250 h showed even more remarkable decrease to 7.5. There seems to be a reasonable cause for this, since the laboratory was evacuated for safety reasons at the time of the ph measurement for the sample and the vial stayed open for two hours. This is rather long time compared to approximately 5 minutes for all the other samples. It seems that the ph affected the distribution of the Eu considerably, since the sample with remarkable ph decrease shows also different Eu distribution (see Figure 7.9). The ph decrease is probably because of intrusion of carbon dioxide. The extent of colloidal fraction in this sample was just around 10 % when in the parallel sample, it was over 40 %. This could be because of the ph effect. In the other hand, the colloidbound Eu fractions of all the samples started to decrease with the both granite mesh sizes from the first sample but the ph values were not decreasing correspondingly. There seems to be other mechanisms affecting the Eu distribution besides the ph. This could be for example instable colloids. The sample solutions were changed to fresh GGW without colloids or Eu after the sample measurements to examine the possible desorption of Eu from granite to solution. The liquid phase was analysed for Eu but only low concentrations were detected on these samples. The activities in the solutions of desorption samples were % ( mm) and % (1-2 mm) and no time dependency was found, e.g. all the samples showed similar Eu concentrations. 83

86 Figure The ph evolution of batch samples with Eu, Ni-montmorillonite and granite mesh sizes of mm (green diamonds) and 1 2 mm (red squares). Two samples were taken for each point and individual markers are shown for each of these. The red line shows the initial ph before contact with granite. Colloid concentration (Figure 7.12) and particle size (Figure 7.13) of the Nimontmorillonite batch samples were analyzed with LIBD and ICP-OES to ensure the colloid stability. The yellow diamonds represent the results of LIBD measurements and the blue circles the results of ICP-OES for the uncentrifuged samples. The lines in figures show the measurements of desorption (green line) and centrifuged (red line) samples, as well as the calculated values for the initial solution (purple line). LIBD measurements were done twice. The first measurement series (yellow diamonds) suggested that the colloids were stable during the experiment. One of the first samples showed lower colloid concentration but the other concentrations were consistent with the initial concentration. On the other hand, the second measurements (not shown in the graphs) conducted less than a week later showed much lower colloid concentration (<5 mg/l) and smaller particle size. The small number of samples makes it difficult to identify whether the different results are because of failed measurements or the colloid instability. The ICP-OES measurements showed correct concentration for the stock solution (25 mg/l), but the colloid concentrations of samples varied from the values determined with LIBD. The values were somewhat lower with ICP-OES but the concentrations were higher 84

87 Z-average [nm] Colloid concentration [mg/l] for the samples taken later. ICP-OES results showed colloid concentrations between 10 and 20 mg/l. No clear reason was found for the difference between the different samples, but this could be related to the stability of the samples. The LIBD measurements done for the centrifuged or desorption samples did not show any particles. This indicates that the separation process by centrifuging was successful and no colloids were released from the granitic material. 100 Tracer solution ,1 0, t [h] Non-centrifuged (LIBD) Non-centrifuged (ICP-OES) Desorption samples (LIBD) Centrifuged samples (LIBD) Figure Ni-montmorillonite colloid concentrations as the function of time measured with LIBD (yellow) and ICP-OES (blue circles). Purple line stand for the initial colloid concentration before the contact with granite, red line stand for the average concentration of centrifuged samples and green for average concentration of desorption samples Tracer Suspension 10 Non-centrifuged 1. measurement Desorption samples Centrifuged samples t [h] Figure Ni-montmorillonite colloid concentrations as the function of time measured with LIBD (yellow). Purple line stand for the initial colloid size before the contact with granite, red line stand for the concentration of centrifuged sample and green for concentration of desorption sample. 85

88 A batch experiment with 152 Eu and Kuru grey granite ( mm) was done also with different colloids, which composed of Nanocor PGN montmorillonite. Results showed similar behaviour as the previous experiments with Nimontmorillonite although the fraction of colloid-bound Eu decreased slower and the ph values were somewhat lower. The fraction of Eu attached to the colloids (see Figure 7.14) showed descending tendency. After 500 hours, 60 % of Eu was attached to the colloids with Nanocor PGN. For the comparison, less than 50 % of Eu was attached to the Ni-montmorillonite colloids after 200 hours with the same granite mesh size (see Figure 7.9). ph of the samples varied between 7.5 and 8 during the experiments (Figure 7.15) which is slightly lower compared to the ph values of Ni-montmorillonite samples which were mostly higher than 8.5 (Figure 7.11). The colloid particles were analysed with PCS instead of LIBD. The particle size grew from 220 nm to 290 nm. The latter was measured two months after the experiments. Taking the time into account, it was not such a remarkable increase. No colloids were detected in centrifuged samples, which proves that the separation was successful. The colloid-bound fraction of Eu in solution without granite was measured in the beginning and in the end of the experiment (purple line in Figure 7.14). The desorption of Eu was also visible there as the colloid-bound fraction decreased, although it did not reach as low as in the samples including the granite. 86

89 152 Eu c/c0 ph 1 0,9 0,8 0,7 0,6 0,5 0,4 Eu attached on colloids before contact with granite Colloids Granite 0,3 0,2 Dissolved 0,1 0 0,00 200,00 400,00 600,00 t [h] Figure Proportional 152 Eu distribution in GGW solution (blue diamonds), on Nanocor PGN montmorillonite colloids (green triangles) and granite of mesh size mm (red squares). The purple line shows the percentage of Eu attached on colloids before the contact with granite. 10 9,5 9 8,5 8 Tracer solution at the preparation 0,1-0,5 mm 7, t [h] Figure The ph evolution of batch samples with Eu, Nanocor PGN montmorillonite colloids and granite mesh sizes of mm (blue diamonds). Two samples were taken for each point and individual markers are shown for each of these. The red line shows the initial ph before contact with granite. 87

90 7.3. Block-scale experiments on radionuclide migration The main objective of this work was to study the sorption behaviour of montmorillonite colloids and their migration in a natural fracture located inside a granitic block. Totally 18 different experiments were conducted with different non-sorbing and sorbing tracers. The summary of the different experiments is given in Table 7-3 and the results are represented in detail in the following section. 88

91 Fluor Fluoresceine 6000 ppm 50 µl AGA Amino-G 9400 ppm 50 µl AGA Amino-G 9400 ppm 50 µl AGA Amino-G 9400 ppm 50 µl 29 n.d AGA Amino-G 9400 ppm 50 µl AGA Amino-G 11 ppm 58.4 ml AGA Amino-G 11 ppm 9.2 ml Notes id Collection Tracer Concentration Injected Tracer Collection Volume Recovery [%] Flow velocity [ul/min] velocity [ul/min] time/sample [min] Fluor60116 Fluoresceine 6000 ppm 50 µl n.d. a n.d. 120 AGA Amino-G 140 ppm 9.4 ml Block Block H, 36 Cl 4.37 kbq/ml 4.18 kbq/ml 9.27 ml H 4.32 kbq/ml 9.98 ml Block H 4.32 kbq/ml 9.45 ml Block Amino-G 152 Eu 179 ppm kbq/ml 9.44 ml n.d. 60 Block Amino-G, 152 Eu 187 ppm kbq/ml 9.76 ml n.d. 60 Only samples from channel A Block Amino-G, 85 Sr a not determined 182 ppm 1.56 kbq/ml 9.44 ml 18 0 measured Short sample collection

92 Absorbance The dye Amino-G was used as a conservative tracer in most of the block experiments. The concentrations of samples were calculated using standard series with varying concentration of the coloring agent and different standard series were made for the UV-VIS by diluting from Amino-G stock solution. Two different laser wavelengths were used: 310 and 250 nm. Different standard series were prepared for Amino-G with Allard, GGW, MilliQ and with laser wavelengths of 250 and 310 nm. No remarkable difference was found between different water solutions. 250 nm showed better intensity compared to 310 nm, which was the reason it was used for the samples from the block experiment. The standard series of Amino-G in MilliQ measured with 250 nm laser is shown in Figure The measured absorbance values of the standard samples and corresponding concentrations were plotted with MS Excel software and a linear fit to the data points was made. The dye concentrations of samples from the block experiments were calculated using the obtained line equation shown on the chart. 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 y = 41366x + 0,0013 R² = 0, , , , , , c [mol/l] Figure The measured absorbances for Amino-G at the wavelength of 250 nm in MilliQ as the function of Amino-G concentration. The fitting was used to calculate the Amino-G concentrations of measured samples. The first block scale tests were performed using a non-sorbing tracer Amino-G. The injection was done from one location with a 50 l injection loop and the sampling was done from two collection channels A and B. Obtained breakthrough curves had large fluctuation between the different experiments, especially for the channel A (see Figure 7.18). The fluctuation was probably caused by non-repeatable injection via the injection loop. That caused fluctuation in the actual injected volume which caused differences between the experiments. 90

93 1 c/c 0 0,1 0,01 0,001 0,0001 0, , AGA A AGA A AGA A AGA A AGA A Fluor. A , Time from injection [min] 1 c/c 0 0,1 0,01 0,001 0,0001 0, , AGA B AGA B AGA B AGA B AGA B Fluor. B , Time from injection [min] Figure The breakthrough curves for non-sorbing tracers obtained with injection via a 50 l injection loop. The upper graph represents the breakthrough in channel A and the lower graph in channel B. One experiment was done with fluoresceine (dark blue crosses) and the rest five experiments with Amino-G (AGA). The code in the legend refer to the starting date of the experiment. 91

94 The injection system was changed because of the non-repeatable breakthrough curves by completely removing the injection loop and instead injecting the tracer solution directly from a vial to the hose leading inside the rock fracture. Besides Amino-G, also radioactive non-sorbing tracers 3 H and 36 Cl were used in some of the experiments to ensure that non-repeatability was not caused by the behaviour of Amino-G. The results of experiments conducted after the change of the injection system are shown for A and B channels in Figure 7.19 and Figure 7.20, respectively. Although some variation still existed, the breakthrough curves showed much better reproducibility between the different experiments with the new injection. Also, it seemed that the Amino-G, 3 H and 36 Cl show similar behaviour within the experimental conditions. c/c 0 1 0,1 0,01 0,001 0,0001 0, , AGA AGA Cl H H H AGA AGA , Time from injection [min] Figure The breakthrough curves of different non-sorbing tracers for channel A obtained with a direct injection. The legend points the used tracer and the date of experiment. 92

95 c/c0 1 0,1 0,01 0,001 0,0001 0, , AGA AGA Cl H H H AGA AGA , Time from injection [min] Figure The breakthrough curves of different non-sorbing tracers for channel B obtained with a direct injection. The legend points the used tracer and the date of experiment. Initially the block was equilibrated with Allard water simulant. The results of the colloid stability experiments however indicated the instability of Ni-montmorillonite in this type of water. Therefore, the groundwater composition had to be changed, which was done by first rinsing the fracture with MilliQ using a high injection velocity and then changing the water to GGW. The changes of flow field were analysed by repeating the experiment with 3 H before and after the water change. The breakthrough curves from the two experiments are shown separately in more detail for the two channels in Figure 7.21 and Figure The red cross symbols stand for the experiment done before the water change and the purple cross symbols for the experiment after it. The breakthrough curves obtained from the tests with 3 H (green cross) and 36 Cl (blue cross) before the water change are also shown. All experiments showed similar behaviour which indicates that the flow conditions remained same in both channels. 93

96 0,1 0, Cl-36 c/c 0 0, H H-3 0, H-3 0, Time from injection [min] Figure The breakthrough curves of 3 H and 36 Cl in channel A of the block. The experiment represented with purple symbols was done after changing the used water to GGW, while the rest was done before it. The codes in the legend refer to the date of experiment followed by the used tracer. 0,1 c/c 0 0,01 0,001 0, H Cl H H- 3 0, Time from injection [min] Figure The breakthrough curves of 3 H and 36 Cl in channel B of the block. The experiment represented with purple symbols was done after changing the used water to GGW, while the rest was done before it. The codes in the legend refer to the date of experiment followed by the used tracer. 94

97 The experiments with sorbing tracers were performed with Eu and Sr. Amino-G was always also used as an indicator of repeatable injection. The tracer solutions were prepared by equilibrating Ni-montmorillonite colloids, Amino-G and a sorbing tracer for approximately one day. Then solution was injected inside the fracture and the collected samples were put aside for the radioactivity and colloid analysis. Amino-G showed a repeatable breakthrough in the both experiments. Instead, no Eu or Sr breakthrough were detected with the experiments ran with montmorillonite colloids (c = 40 mg/l). The samples were measured with LIBD to see if colloid breakthrough occurred. The particle size determinations are shown in Figure 7.23 and the colloid concentrations in Figure 7.24, where also the breakthrough curve of Amino-G is added to graph. PCS was also used in the determination of colloid concentration but the concentrations were too low for the method. The results of LIBD measurement indicated that no colloid breakthrough occurred. Although some particles existed, it seems that they were not the initially injected montmorillonite colloids since the particle size was way too small: the initially injected particles were between 200 and 300 nm but the LIBD measurements showed particles with diameters below 100 nm. To furthermore analyse the particle composition ICP-MS measurements were done for some of the samples. Low and stable concentrations of Ni were detected which corresponded for Ni-montmorillonite concentration of 0.10 ± 0.01 mg/l. The concentrations were too low for the colloid breakthrough but could originate from the montmorillonite dissolution for example. The deposition of montmorillonite colloids inside the fracture explains why Eu was not detected. Trivalent Eu sorbs strongly to mineral surfaces and is retained strongly if not attached to colloids. The reason for the lack of Sr breakthrough remains an open question because Sr should act as a moderately sorbing tracer, which should migrate slower compared to non-sorbing tracer. Instead of slightly retarded behaviour, Sr was not detected at all. 95

98 c/c0 Particle Size d [nm] Tracer solution Block experiment A Time [h] Figure Particle sizes of the samples collected from block experiment done with Nimontmorillonite colloids and Eu. Particle sizes were below 100 nm (yellow diamonds), while the originally injected particles were between 200 and 300 nm (red line). 0,1000 0,0100 AGA A ,0010 Block experiment A , Time [min] Figure Breakthrough curves of Ni-montmorillonite colloids (yellow squares) and Amino-G (blue crosses) of run done with 152 Eu tracer and Ni-montmorillonite colloids. Amino-G breakthrough (blue cross) was obtained but no colloids or Eu were detected. 96